This week’s PMF Playbook comes from my episode with Monzy Merza, Founder and CEO of Crogl.

Monzy is one of those founders who didn’t discover a problem from the outside. He lived it. He spent more than 25 years in cybersecurity, including time at Splunk, Databricks, and inside the cybersecurity team at HSBC, before starting Crogl.

What made the conversation valuable wasn’t just the product. It was the way Monzy thinks about founder-market fit, customer listening, and the difference between hearing a problem and truly understanding the physics of that problem.

Let me walk you through what stood out.

Founder-market fit: living the problem before building the solution

A lot of founders today build products in industries they’ve never worked in.

Sometimes that works. But in complex markets like cybersecurity, surface-level understanding is dangerous. The easy assumptions are usually wrong.

Monzy’s journey to Crogl started with a pattern he kept hearing from customers at Splunk and Databricks: security teams were never going to put all their data in one place.

That sounds simple. But it challenged years of conventional wisdom.

The standard pitch in data and security was: centralize your data, put it in one lake, and then analytics will solve the problem. But customers were saying something different. Their data was already spread across S3, Databricks, Splunk, log analytics platforms, endpoint tools, cloud services, and countless other systems.

And they were not going to move it all.

The PMF lesson is this: sometimes the market tells you the answer repeatedly, but you only hear it once you stop trying to force it into your existing worldview.

Listening to customers is not the same as hearing customers

One of the strongest parts of the conversation was Monzy’s definition of listening.

Most people say “listen to the customer.” But what does that actually mean?

Monzy described it as understanding the physics of the customer’s work. Not just the words they use. Not just the complaint. The actual day-to-day motion of the job.

If a customer says, “We want more value from our data,” the lazy interpretation is: great, here’s a better data platform.

The deeper interpretation is: what does value mean? Why are you not getting it today? What outcome would prove that value has been created? What work has to happen for that outcome to exist?

That is where the real problem lives.

In cybersecurity, the obvious phrase everyone uses is alert fatigue. Too many alerts, not enough analysts, too much noise.

But when Monzy dug deeper, he found the real problems underneath:

Analysts had domain knowledge gaps. They had tool competency gaps. And teams had collaboration gaps, where one analyst’s work did not easily compound into another analyst’s work.

That is a very different problem than “too many alerts.”

And that distinction matters.

If you solve the tagline, you build another feature. If you solve the physics, you build a company.

The aha moment: “we are never putting all our data in one place”

The moment that led to Crogl happened during a Databricks customer call.

The customer said they were never going to put all their data in one place.

Monzy had heard versions of that before, but this time it landed differently. It triggered a realization: if security data is permanently distributed, then the old model of centralizing everything before investigation is fundamentally broken.

That insight did not immediately become a pitch deck.

Instead, Monzy spent weeks challenging himself. Has this already been solved? Why had federated search not fixed it? Why had decades of tooling still left security analysts overwhelmed? What would a person on the keyboard actually need to do the job properly?

That discipline matters.

Aha moments are rarely enough. The founder still has to interrogate the insight until it either breaks or becomes stronger.

Why Monzy went back inside the problem

Before founding Crogl, Monzy did something most founders would never do.

He went back into the operator seat.

He joined the cybersecurity team at HSBC because he wanted to experience the problem directly in one of the most complex environments possible.

That decision says a lot about how he thinks.

HSBC had the resources. It had the talent. It had the tools. If even an organization like that still struggled with alert investigation, data sprawl, and analyst constraints, then the problem was not simply budget or hiring.

It was structural.

This is an important founder lesson: when you remove the obvious excuses and the problem still exists, you may have found something worth building.

Crogl: what the company actually does

Crogl works on alerts.

That was Monzy’s simplest definition.

In a security operations center, analysts can receive hundreds or thousands of alerts. Most are not meaningful. But one missed alert can become the root of a breach, a regulatory issue, a financial loss, or a CISO losing their job.

The conventional answer has been to filter alerts down.

Monzy calls that the diminishing view of the world: the idea that everything should somehow get smaller. Fewer alerts, less noise, fewer things to inspect.

Crogl takes the opposite view.

The world gets bigger. Data grows. Computing grows. Attack surfaces grow. Alerts grow.

So instead of assuming fewer alerts, Crogl investigates every alert with depth. It connects across the tools and data an organization already has, helps analysts understand whether there is evidence of threat, and documents the work automatically.

The analyst does not have to remember every schema, every query language, every tool convention, or every organizational nuance.

Crogl brings that context forward.

The PMF lesson here is clear: a great product does not always remove work. Sometimes it removes the wrong work so humans can focus on the judgment they were hired for.

The real customer: the practitioner who carries the pain

Crogl sells to security leaders, but it is built for practitioners.

That distinction matters.

The buyer might be a VP of security engineering, a security leader, or in some cases the CISO. But the user is the analyst, the threat hunter, the incident responder - the person actually doing the work.

Monzy’s pride in Crogl is tied directly to that community.

In ten years, he wants Crogl to have made analysts look like the heroes they already are. He wants their work to be captured, reused, improved, and recognized. He wants the domain knowledge gap, tool competency gap, and collaboration gap to be materially smaller.

That is a powerful founder-market fit signal.

The company is not just serving a budget line. It is serving a community.

The PMF lesson: boring problems can be the most valuable

One of Monzy’s strongest points was that many of the easy problems have already been solved.

What remains are often the boring ones.

The messy ones. The operational ones. The problems buried inside workflows that outsiders underestimate because they look unglamorous from a distance.

But those are often the highest-value problems.

In security operations, the work is not sexy. It is repetitive, detailed, high-stakes, and unforgiving. But if you can solve it, or even materially improve it, the value is enormous.

This is where a lot of founders get founder-market fit wrong.

It is not just “I know the space.” It is not just “I worked in this industry.” It is the willingness to get dirty, respect the complexity, and serve the people doing the work.

AI in security: not replacing people, but multiplying them

Monzy was clear on one thing: broad claims about replacing humans usually reveal shallow understanding.

In security, the work is too nuanced to reduce to a slogan.

Crogl is not valuable because it says, “you no longer need security analysts.” It is valuable because it helps analysts operate at a much higher level.

It investigates alerts. It documents work. It learns from analyst input. It helps analysts avoid being blocked by tool knowledge or data sprawl.

That is a much more realistic AI wedge.

Not replacement.

Amplification.

And in enterprise markets, amplification often wins because it respects how work actually happens.

The deeper takeaway: PMF starts with service

The theme I kept coming back to in this episode was service.

Monzy did not describe Crogl as a clever AI product. He described it as a way to serve a community he understands deeply.

That changes how you build.

You ask better questions. You avoid arrogant assumptions. You do not dismiss complexity. You do not build a magic wand and then go looking for a market.

You start with the people doing the work.

Then you build around their reality.

Closing thought

If I compress the entire episode into one sentence, it’s this:

PMF starts when a founder understands the physics of a customer’s work deeply enough to solve the problem beneath the obvious problem.

Crogl is not just solving alert fatigue.

It is solving the knowledge, tooling, and collaboration gaps that make security operations so difficult in the first place.

That is the kind of insight that only comes from listening properly, living the problem, and caring enough about the practitioner to build something that actually helps.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Evan Silberhorn. Evan has lived across multiple company-building worlds: operator, founder, consultant, BCG Digital Ventures, and now private equity portfolio operations.

What made this conversation valuable was the contrast between venture capital and private equity. Not as abstract finance categories, but as two very different operating systems for value creation.

Let me walk you through what stood out.

VC versus PE: growth versus durability

Evan’s simplest distinction was also the clearest.

Venture capital is about growth. You invest early, often before product-market fit is fully proven, and you put capital behind founders who might be able to capture a massive market. The goal is speed, iteration, market capture, and scale.

Private equity is different. It is not just asking, “Can this company grow?” It is asking, “Can this company grow while improving the bottom line?”

That’s where EBITDA becomes the language of the room. In PE, value creation is not just top-line expansion. It is top-line growth plus cost discipline, process improvement, operational maturity, and margin expansion.

The PMF lesson here is important: venture is often about finding and scaling PMF; private equity is about making PMF more durable, measurable, and profitable.

The PE lens: no BS, just value creation

One phrase from Evan captured the private equity mindset perfectly: there is no BS.

When a PE firm buys or invests in a company, everything starts with a value creation plan. That plan becomes the operating map for the hold period. Product roadmap, sales strategy, cost structure, AI initiatives, hiring, organizational design - everything needs to connect back to how value will be created.

That’s a very different rhythm from the early-stage startup world, where ambiguity is often part of the journey.

In PE, ambiguity has a cost. You are operating against a clock. You need to show progress quarter after quarter, and the company has to become more valuable within a defined time window.

The PMF lesson: once a business reaches scale, storytelling alone is not enough. The story has to become an operating plan.

The real PE clock is shorter than founders think

One of the most useful parts of the conversation was Evan’s breakdown of the private equity timeline.

People often hear “five-year hold” and assume there are five full years to create value. But Evan explained that the real window is much shorter.

The first six to eight months are often spent aligning on the value creation plan, understanding the company deeply, improving visibility, and setting up the operating system. Then, by year three and a half or four, the firm is already thinking about exit positioning.

That means the true value creation window may only be from month eight to around year three and a half.

That changes the whole game.

You don’t have unlimited time to modernize the stack, fix the operating model, rebuild the product motion, integrate acquisitions, and improve growth. You have to prioritize ruthlessly.

The PMF lesson: time is not just runway. Time is strategic leverage. If you waste the first year, you may have already lost the hold.

Why PE cares so much about operations

Evan made the point that many PE-backed companies are not broken. They are often proven companies that have grown through sales motion, founder energy, or acquisition.

But they may be operationally immature.

They may have multiple product lines that do not fully connect. They may have different systems stitched together from acquisitions. They may have a sales-led culture that never fully evolved into a product-led or customer-led operating model.

That is where PE sees opportunity.

The value comes from bringing maturity: better reporting, better processes, clearer product strategy, stronger go-to-market alignment, cleaner cost structure, and more disciplined execution.

The PMF lesson: growth creates complexity. If you don’t professionalize the operating system, the complexity eventually eats the growth.

AI in PE: efficiency first, transformation second

Evan’s view on AI was pragmatic.

In the venture world, AI today can feel like a land grab. Capital is being poured into infrastructure, talent, and capability-building. Companies are paying extraordinary compensation because they are terrified of falling behind.

But in private equity, the AI conversation is becoming much more focused on efficiency.

Where can AI reduce cost? Where can it improve workflows? Where can it help the company do more without endlessly adding headcount? Where can it drive measurable business value?

That’s a different adoption pattern from the hype cycle.

The PMF lesson: AI will not be judged forever on pilots and demos. Eventually, it will be judged on business outcomes.

Founder autonomy changes after PE

One of the sharper points in the episode was what happens when a founder takes private equity money.

Evan was direct: the PE firm becomes your boss.

That does not mean the founder loses all control overnight. But it does mean the founder now reports into a very different accountability structure. Quarterly performance matters. EBITDA matters. The value creation plan matters.

If the founder can scale with the company, great. If not, PE firms may decide that the next phase requires different leadership.

That can be difficult because founder-led companies often have deep emotional and cultural attachment to the founder. Removing a founder can disrupt the organization. But if the company is missing targets and the next stage requires a different operating muscle, the conversation becomes unavoidable.

The PMF lesson: the person who creates PMF is not always the person who scales PMF through the next phase.

East Coast versus West Coast capital

I loved Evan’s framing of East Coast versus West Coast investing.

Silicon Valley is more comfortable taking big swings on founders, ideas, and category-defining outcomes. It is more willing to believe in a future that does not yet exist.

The East Coast mindset, especially in private equity, is more measured. It is more numbers-oriented, more durability-focused, and more anchored in proven business models.

Neither is right or wrong. They are simply different forms of capital with different expectations.

Venture asks: how big can this become?

Private equity asks: how durable, profitable, and operationally excellent can this become?

The PMF lesson: the type of capital you take shapes the company you become.

The support you get: networks versus playbooks

Another useful contrast was the support founders receive.

From venture capital, founders often get access to networks: other founders, early-stage operators, talent, customers, and advisors who can help them find PMF and scale quickly.

From private equity, companies get playbooks: former operators, proven processes, operational benchmarks, cost discipline, reporting systems, and repeatable methods for improving performance.

VC helps you move faster into uncertainty.

PE helps you operate better inside a proven model.

The PMF lesson: founders should not just ask, “Who will give me money?” They should ask, “What kind of operating system does this capital bring with it?”

Closing thought

If I compress the entire episode into one sentence, it’s this:

PMF may begin with growth, but long-term value comes from turning that growth into a durable, disciplined, measurable operating system.

That is the bridge between venture and private equity. Venture funds the possibility. Private equity tests the machinery.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Sandy Climan. Sandy is one of those rare operators whose career only makes sense when you zoom out wide enough. He has spent decades at the intersection of entertainment, media, venture, and technology, watching industries collide long before most people even realized they were moving toward each other.

What made this conversation unusually valuable wasn’t a single tactic or a simple framework. It was the way Sandy described convergence itself: how worlds that once looked separate eventually become inseparable, how consumer behavior shifts before institutions fully understand it, and how the winners are usually the ones who see not just the product in front of them, but the platform underneath it.

Let me walk you through what stood out.

The false start: what Silicon Valley and Hollywood got wrong the first time.
Sandy’s answer to the original question was subtle, but important: the first wave of Hollywood-meets-Silicon-Valley didn’t fail because the idea of convergence was wrong. It failed because the infrastructure wasn’t ready.

In the 1990s, the vision was already there. People could see that technology and storytelling were going to come together. CD-ROMs, multimedia, interactive entertainment, early digital formats, all of it felt like the future. But the actual consumer experience lagged too far behind the ambition. Uploading and downloading was too slow. Interactivity was primitive. Distribution wasn’t fluid yet. The internet had not fully become the connective tissue.

That is such an important PMF lesson. Sometimes founders are directionally right and still commercially early. The market does not reward prophecy on its own. It rewards timing, infrastructure readiness, and behavioral adoption. If even one of those is missing, the vision looks wrong until suddenly it doesn’t.

The deeper point is that premature convergence can look like failure when it is really just unmet conditions.

The real bridge was not content. It was the internet.
One of the strongest ideas from the conversation was that the internet became the API between Silicon Valley and Hollywood.

Before that, technology and entertainment operated as neighboring countries with different languages, different incentives, and different rhythms. Tech was building tools. Hollywood was building stories. The CD-ROM era hinted at overlap, but it was the internet that finally made the connection real.

That shift mattered because it changed not only creation, but distribution, accessibility, and consumer expectations. Once content could move fluidly through connected networks, entertainment stopped being bound to theaters, schedules, and physical formats. It became software-like: searchable, streamable, personalized, and increasingly global.

That is where the PMF insight gets interesting. Entire categories unlock when something becomes programmable. The internet didn’t just improve media distribution. It changed the unit economics, the speed of experimentation, and the reach of storytelling itself. Once that happened, Hollywood was no longer just an art business. It became a platform business too.

The cultural gap: why Silicon Valley and Hollywood kept misunderstanding each other.
Sandy described the difference between Hollywood and Silicon Valley in a way I haven’t heard put so clearly before.

In Silicon Valley, people will often give almost anyone fifteen minutes if the idea is interesting enough. The culture is practical, iterative, and biased toward testing. In Hollywood, the barrier to entry traditionally has been relational. Trust comes first. Business often follows later. In one world, transaction can lead to relationship. In the other, relationship often needs to come before transaction.

That difference created enormous friction in the early attempts to connect the two industries. Tech people found Hollywood inefficient. Hollywood people found tech naive. Both sides underestimated how much of their own operating system they were taking for granted.

This is a PMF lesson that goes beyond products. Sometimes what looks like market resistance is actually cultural mismatch. You are not just selling a new capability. You are introducing a new logic for how decisions get made. If the customer’s system of trust, status, or relationship formation is different from yours, your product can be correct and still fail to land.

Founders often talk about fit as if it is only about user pain. But there is also process fit, cultural fit, and timing fit. Ignore those and the market will feel colder than it really is.

From audiences to communities: the biggest shift in modern media.
Sandy said something that I think is one of the most important ideas in this whole conversation: audiences are becoming communities.

That sounds simple, but it changes everything.

The old entertainment model was built around broad distribution and probabilistic demand. Make something, market it widely, hope it hits. The new model is increasingly about identifying, aggregating, and serving a specific community that already shares a value system, identity, or obsession. In that world, success becomes less about mass guessing and more about precise resonance.

That is why his Angel Studios example is so powerful. The product is not just the film. The product is participation. The members are not passive viewers. They are part of a loop. They vote, engage, share, and identify with the larger mission. That changes retention, monetization, and distribution all at once.

For founders, this is a huge PMF principle. Product-market fit becomes much stronger when your market stops behaving like an audience and starts behaving like a community. Communities do more than consume. They reinforce identity, generate word of mouth, increase switching costs, and create emotional durability around the product.

The best companies today are not simply acquiring users. They are organizing believers.

What Netflix really understood.
When we think about Netflix, we usually talk about streaming. Sandy’s framing was more nuanced. Netflix won twice, in two different ways.

First, it used new technology and a new path of distribution to beat Blockbuster. DVDs were easier to ship than videocassettes, and mailing was a more efficient distribution channel than physical retail. That alone was enough to create discontinuity.

Then Netflix did the harder thing. It disrupted itself before someone else could. It recognized that DVDs-by-mail were not the endpoint. Streaming was. Even when bandwidth was imperfect and the experience was still emerging, it chose to move early.

That is one of the clearest examples of durable PMF thinking. Product-market fit is not a resting place. It is a position you keep having to abandon in order to stay ahead. The version of the company that first wins is often not the version that deserves to keep winning.

Founders love to talk about finding PMF. The harder question is whether they are willing to cannibalize the thing that got them there.

Why so many successful companies still lose.
Sandy’s explanation here was one of my favorite parts of the episode. He argued that many companies fail not because they stop being smart, but because success creates blind spots.

They become too close to their own trees to see the forest. They fall in love with their own invention. They look for evidence that the current model still works instead of asking what would replace it. They confuse inside opinion with outside fact.

That distinction was brilliant: opinion exists on the inside, and fact can only exist on the outside.

That is an incredibly useful way to think about product-market fit. Internally, companies can convince themselves of almost anything. They can rationalize roadmap decisions, defend complexity, and explain away signals that no longer support the story. But the market is merciless. Customers do not care how elegant your internal logic is. They reveal the truth through behavior.

This is why PMF is never a founder’s declaration. It is always a market verdict.

AI, concierges, and the new interface layer.
Toward the end of the conversation, Sandy pushed into a broader frame that felt especially timely: we are moving toward a world of algorithmically driven concierges.

That does not just mean better recommendations. It means the interface layer for daily life gets restructured. Information, entertainment, commerce, health, finance, and communication increasingly move toward systems that do not just present options, but guide, anticipate, and act.

In his words, we all want an easy button.

That idea matters because it changes how culture gets made and how products get discovered. If algorithmic systems mediate what we watch, buy, read, and care about, then the battle shifts from creating content or tools in isolation to understanding how those systems shape access and behavior.

The PMF implication is enormous. In the next era, great products may not just need to win with the user. They may need to win with the concierge layer between the user and the world. That means context, relevance, trust, and adaptability become even more important. It also means distribution becomes less static and more dynamic than ever before.

Why Sandy is not afraid of AI.
I liked Sandy’s answer on this because it was measured. He does not dismiss AI, and he does not romanticize it either. He sees it as a tool, and in many cases, a powerful one. He believes it will become an extraordinary inspirational and editing engine. But he is still anchored in the idea that true creative spark comes from human vision.

That perspective matters. Every new technology wave tempts people into one of two bad habits: either assuming the new tool changes nothing, or assuming it changes everything. Both are lazy. The real work is understanding exactly what changes, what stays human, and what becomes newly possible when the tool enters the system.

This is true in media and in startups. AI will absolutely reshape workflows, lower production barriers, and alter the economics of creation. But that does not eliminate the value of taste, instinct, or narrative judgment. In fact, as tools become more abundant, those human qualities probably matter even more.

The stories that endure are still going to require someone who can see what others do not.

The broader takeaway: PMF is increasingly about convergence.
If I compress the entire episode into one sentence, it is this:

The next wave of product-market fit will belong to the people who understand not just products or industries in isolation, but the convergence layer between technology, behavior, community, and culture.

That is what Sandy sees so clearly. Hollywood and Silicon Valley were once separate worlds. Then the internet connected them. Now AI, data, platforms, and algorithmic distribution are pushing them into an even deeper merger. And the winners will not be the ones who cling to the old categories. They will be the ones who understand what new human behavior becomes possible when the categories collapse.

That is where the future gets built.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Aaron Fulkerson. Aaron is the CEO of Opaque, a company building what may become one of the most important infrastructure layers of the AI era: confidential AI. But what made the conversation unusually valuable wasn’t just the technology. It was the deeper idea underneath it all: trust.

Aaron has spent a large part of his life thinking about trust, not as a vague virtue or a soft cultural theme, but as a foundational system. Something that determines whether teams function, whether societies remain stable, and whether new technologies create prosperity or chaos.

What made this conversation stand out is that it connected trust, AI, platform shifts, enterprise value, and human behavior into one continuous PMF story. You can’t separate them. If AI is the next major platform shift, then trust is the infrastructure that decides whether that shift creates durable value or systemic fragility.

Let me walk you through what stood out.

The starting point: trust is not abstract, it is structural
Aaron defines trust through three pillars: caring, consistency, and competency.

Caring means I believe you have my interests in mind. Consistency means I know how you will show up tomorrow because I’ve seen how you show up today. Competency means I believe you can actually do what you say you will do.

That framing stayed with me because it gives trust operational meaning. It moves trust out of the world of slogans and into the world of systems. It also explains why trust is so easy to talk about and so hard to build. You do not get it by declaring it. You get it by repeatedly demonstrating care, reliability, and capability over time.

The PMF lesson buried in that is simple: trust is not a brand asset layered on top of product-market fit. It is often part of the product itself. In many categories, especially the ones touching data, workflows, or critical decisions, trust is inseparable from adoption. If the buyer does not trust the system, PMF never fully forms no matter how impressive the demo looks.

The deeper question: every platform shift needs a trust layer upgrade. One of Aaron’s most important ideas is that every major technology platform shift requires a corresponding trust upgrade.

The internet is the clearest example. Early on, it was a network of pages and information. But it could not become a global commerce platform until the trust layer improved. Encryption in transit, HTTPS, SSL, TLS all of that mattered because it gave people confidence that transactions could happen safely. Without that upgrade, the internet remains interesting but limited.

Aaron’s argument is that AI is now going through the same moment.

The difference is that the trust problem is no longer just about data in transit. It is about runtime verifiability. It is about what happens when autonomous or semi-autonomous systems operate at machine speed, touching sensitive data, making decisions, chaining actions across systems, and interacting with enterprise workflows in ways that humans alone never could.

That distinction matters. A human being, even a bad actor, is constrained by time, energy, and coordination. An agentic system is not. And once you move from software as a passive tool to software as an active actor, the trust model has to change with it.

The PMF takeaway here is big: if AI is the platform shift, then trust is not a feature request. It is the enabling condition for the market to exist at scale.

Why enterprise AI adoption stalls without trust. There is a temptation in AI to think the biggest product question is intelligence. How good is the model? How fast is the output? How magical is the interface?

Aaron’s view is that for enterprises, that is not the gating factor. The gating factor is whether they can use their most valuable data without exposing themselves.

That is where his examples became especially sharp. A company may not be worried that a foundation model is literally training on its data in the most direct sense. The more subtle concern is the metadata, the workflows, the ways and means by which people use the model in conjunction with proprietary information. That interaction pattern itself can become valuable intelligence.

And that changes the equation.

The PMF lesson here is uncomfortable but important: in the AI era, product value may increasingly come not from owning the model, but from protecting the environment in which the model is used. The company that makes AI usable with proprietary data may create more durable value than the company that simply makes AI impressive.

Opaque’s wedge: confidential AI as the HTTPS moment for enterprise AI. This is where Opaque becomes strategically interesting.

Opaque is not trying to win by being another application layer company wrapped around a model. It is building around a much deeper problem: how do you let enterprises use AI on sensitive data with verifiable guarantees around privacy, policy enforcement, and confidentiality?

The simplest analogy Aaron gave is that this is the AI equivalent of the internet’s encryption moment. Enterprises want to do things like confidential RAG, summarization across HR, legal, customer, and finance systems, and agentic workflows that pull together multiple data sources. But they cannot do that safely if they are effectively handing over control of their most sensitive data exhaust.

Opaque’s approach is built around confidential computing and verifiable policy enforcement before, during, and after runtime. The details get technical quickly, but the strategic point is clear: it allows companies to use AI without surrendering the underlying crown jewels that make their businesses valuable.

And that creates a very specific PMF pattern. The product is not just solving for “can this AI do the task?” It is solving for “can this AI do the task in a way our business can actually adopt?” Those are very different questions. One creates demos. The other creates budgets.

The platform risk founders should be paying attention to. Another part of the conversation that stuck with me was the broader warning about platform capture.

We have seen this movie before. Toys “R” Us built with Amazon and helped strengthen the platform that eventually disintermediated them. Marketplace operators learn from the activity happening on top of them. The platform gets smarter than the participant.

Aaron’s point is that the same dynamic is emerging again in AI. If the next generation of AI platforms becomes the place where workflows, prompts, interactions, metadata, and downstream customer behavior all accumulate, then the platform does not just provide utility. It begins to absorb the value layer.

That is what makes confidential AI strategically important. It is not just a security layer. It is a bargaining-power layer. It gives enterprises a chance to participate in the AI economy without naively handing the future of their business to the infrastructure provider.

The PMF lesson here is one founders should take seriously: building on a platform is not inherently bad, but building in a way that leaks the source of your long-term defensibility is fatal. PMF that depends on someone else not learning from your best workflows is fragile PMF.

The darker side: trust erosion at machine speed. Toward the middle of the conversation, things got heavier in a way I thought was necessary.

Aaron talked about model poisoning, hidden agendas, and the possibility that AI systems could shape beliefs, manipulate behavior, or amplify subtle influence at a scale that human bad actors simply cannot match. Whether you take every example literally or not, the underlying point is hard to dismiss: systems that operate at machine speed with broad interface access can alter human behavior faster than we are used to defending against.

That is where the conversation moved beyond enterprise software and into society itself.

We talked about people already using AI to respond to personal messages, to mediate emotionally charged communication, and to think on their behalf. There are clearly positive uses for that. But there is also a real cost. The more AI sits between person and person, decision and decision, the more it shapes how we communicate, what we trust, and even how much independent judgment we retain.

The PMF relevance here is subtle but important. Products are not neutral once they scale. They train behavior. They change defaults. They redefine expectations. Founders building in AI are not just shipping utilities. They are shipping interaction patterns that can compound culturally over time.

The human counterweight: empathy, communication, and values. One of the most hopeful parts of the conversation came when Aaron shifted back toward what remains deeply human.

His view is that as AI becomes more pervasive, some human capabilities become more important, not less. Empathy. Communication. Storytelling. The ability to build trust. The ability to connect around shared values.

That also flowed into a second theme he emphasized strongly: the only way to build a high-performance team is on a foundation of trust. And if you do not provide a clear mission and values, people default to fear, ego, and fragmentation.

That part resonated with me a lot because I have seen it inside my own business. Values are useless if they live on a website and nowhere else. They only matter when they shape who you hire, who you promote, what behavior gets rewarded, and what behavior gets filtered out. Otherwise, they are branding. Not culture.

The PMF lesson here is that product-market fit inside a company matters too. A team cannot compound around a product if it cannot compound around a shared way of operating. Internal trust affects external execution.

Bias for action: the founder trait Aaron kept coming back to. Aaron also made a point I think more founders need to hear: people need to develop opinions and test them.

Not endlessly wait for instructions. Not outsource judgment. Not sit in abstraction. Have a view. Try it. Learn. Adjust. Repeat.

He described it as strong opinions, weakly held. That balance matters. You need enough conviction to move, but enough humility to change course when evidence shows up. Without that, you get one of two failure modes: passivity, where no one drives anything forward, or rigidity, where people cling to bad ideas long after reality has moved on.

That is a PMF principle too. PMF is found by iterating against reality. That requires action. And action requires people who are capable of forming a judgment, making a move, and learning fast.

The bigger pattern: AI may commoditize products, but it will magnify values. One of the most interesting closing threads was Aaron’s point that as AI accelerates commoditization, values become even more important.

If products can be built faster, copied faster, and iterated faster, then what becomes a durable point of differentiation? It is not just raw functionality. It is the trust people have in the company, the brand, the mission, and the values embedded in the experience.

That does not mean values replace product quality. It means that in a world of increasing product abundance, values become part of what makes a product meaningful and durable.

In other words, the more AI compresses the cost of creation, the more human trust becomes part of defensibility.

Closing thought:
If I compress the entire episode into one sentence, it’s this:

Every major platform shift creates new capability, but the companies that matter most are the ones that build the trust layer that makes that capability usable at scale.

That is what made this conversation feel so important to me. Aaron is not just talking about AI safety in an abstract sense. He is pointing at a very practical market truth: if enterprises cannot trust AI with their most valuable data and workflows, adoption stalls. And if society cannot rebuild trust while AI scales, the consequences go far beyond enterprise software.

The future of AI will not just be decided by model quality or interface design. It will also be decided by who makes trust operational.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with David Paffenholz, co-founder and CEO of Juicebox. David is building in a category I know intimately: recruiting. That made this conversation especially interesting to me, because it wasn’t abstract. It sat right at the intersection of founder conviction, product design, hiring pain, and market timing.

What made the conversation valuable wasn’t just that David is building recruiting software. It was the deeper story underneath it: how a founder chooses a market, how conviction survives hard fundraising, why speed matters in a zero-sum talent game, and what it really takes to turn an AI workflow into something customers trust enough to use every day.

Let me walk you through what stood out.

Why recruiting is a much bigger market than most people think

One of the first things David said has stuck with me:

Recruiting is going to become even more important than it is today because the competition for talent is only going to increase.

That sounds obvious on the surface, but it’s actually a very sharp PMF insight.

A lot of founders build in markets that feel exciting because the technology is novel. David and his co-founder chose recruiting for a different reason: they believed the underlying pain would compound over time. That is a much better foundation for company-building. They weren’t chasing a trend. They were chasing a pressure that gets worse as the economy becomes more specialized, more competitive, and more talent-constrained.

That matters because PMF is rarely about building something “cool.” It is much more often about identifying a market where urgency keeps rising.

If the pain gets sharper every year, the product has room to deepen. If the pain is temporary, the company ends up fighting for oxygen.

The lesson here is simple: the best markets are not always the loudest ones. They are often the ones where structural pressure is building underneath the surface.

The origin story: PMF often starts with a personal frustration

David didn’t start with a grand theory about HR software. He started with something much more human.

He and his co-founder felt that a lot of the best opportunities they’d had in their own careers were ones they had to create for themselves. That idea led to a bigger question: how do you create more of those opportunities for more people?

That pushed them toward recruiting, and then more specifically into recruiting software.

I think there’s an important PMF principle buried in that. The strongest companies often come from founders who are not just interested in a category, but emotionally bothered by a specific inefficiency inside it.

That doesn’t mean personal pain alone is enough. It isn’t. But it does mean the founder starts with a level of sensitivity to the problem that outsiders usually don’t have.

In David’s case, the initial instinct was not “let’s build AI recruiting software.” It was “there is a broken matching problem here, and talent allocation matters more than people realize.”

That’s a far better starting point.

The first version usually isn’t the company

Another thing I liked about this conversation was how honest David was about the path to Juicebox.

He and his co-founder didn’t emerge fully formed with the perfect idea. They built consumer apps. They tested different concepts. Some didn’t work. One music discovery app got real usage. Then, after YC, they launched what was effectively a talent marketplace and manually lived the workflow themselves.

That part is important.

Before they became a software company, they became very close to the problem. They did the messy work. They learned the workflows. They built internal tooling for themselves. Then the model shift happened, GPT improved, and they realized the software layer was the real opportunity.

That is such a classic PMF pattern: you start with one expression of the problem, but if you stay close enough to the user, the real company reveals itself.

Founders sometimes think changing direction means failure. Most of the time it means learning.

Juicebox resetting revenue to zero when they shifted from marketplace revenue to software revenue is a good reminder that some of the best moves in a company’s life feel backwards in the moment. On paper, you are losing traction. In reality, you may be finally aligning the company with the right long-term model.

Speed is not a nice-to-have in recruiting. It is the product.

One of the strongest ideas in the episode was David’s explanation of why recruiting is a zero-sum game.

Only one company gets to hire a candidate at a given moment. That means speed is not just an efficiency metric. It is an outcome metric.

That changes how you think about PMF.

In a lot of software categories, a better product wins because it is more elegant, cheaper, or more integrated. In recruiting, better often means faster while retaining quality. If your customer gets to the right candidate first, that advantage is material.

That’s what I think Juicebox understands well. It isn’t just trying to create a prettier search experience. It is trying to reduce the time between search intent and candidate discovery without destroying signal quality.

That is a much more powerful positioning than “AI for recruiting.”

The PMF lesson here is that speed only matters if it compounds into a better business outcome. In recruiting, it does. The company that identifies, reaches, and engages the right person first often wins.

So speed here is not cosmetic. It is core.

Signal-to-noise is where real value gets created

This was the part of the conversation that resonated most with me as an operator and recruiter.

The best thing a tool like Juicebox can do is not just return more profiles. It is to improve the signal-to-noise ratio.

That’s where recruiting tools often fail. They give you more data, more candidates, more filters, more screens, more dashboards. But the recruiter still has to do the hard cognitive work of figuring out who actually matters.

What David described, and what I’ve seen in practice, is a product built around reducing that burden.

Not just search.
Not just filters.
Not just profile access.

But actual ranking, interpretation, and narrowing of relevance in a way that starts to feel like the product is doing part of the work for you.

That’s a huge distinction.

A lot of software claims automation. Very little software creates the feeling that your workload has genuinely been reduced. That feeling is where sticky PMF lives.

If the customer finishes the workflow faster but with the same confidence, you’ve created value.
If they finish it faster with better confidence, you may have something exceptional.

Why product-led growth worked here when it rarely works in HR tech

David made a point that I thought was especially sharp: historically, HR tech has not been product-led. It has been sales-led.

That’s true. Most products in the category are top-down, demo-heavy, procurement-heavy, and slow to show value.

Juicebox went the other direction.

Why did that work?

Because the time to value is short.

That’s the real test for PLG. Not whether PLG is fashionable. Not whether founders prefer it. But whether the customer can get to a meaningful “aha” moment quickly enough without a human dragging them there.

In Juicebox’s case, that moment is very tangible. You search. You uncover someone you likely would not have found otherwise. You see quality quickly. You feel leverage immediately.

That is exactly the kind of product experience PLG needs.

There’s a broader lesson here for founders. PLG is not a brand choice. It is a product truth. If your product cannot create trust quickly, PLG will feel forced. If your product can demonstrate obvious value in minutes, PLG can become a real growth engine.

Founders underestimate storytelling at their own expense

One of the most honest parts of the episode was when David said he underestimated the importance of storytelling.

I think a lot of technical founders do.

They assume the product will speak for itself. Eventually, maybe it does. But early on, storytelling is how investors, employees, customers, and future believers understand why the product matters before the metrics are obvious enough to prove it.

David put it well: you eventually earn the right to rely less on story because the traction becomes the story. But until then, you need a compelling narrative bridge between what exists today and what the company is becoming.

That matters in fundraising, obviously. But it matters just as much in hiring and sales.

How do you convince a great engineer to join an early company?
How do you convince a customer to trust a young product?
How do you convince yourself to keep going when the market doesn’t yet fully understand what you’re building?

Story is not fluff. It is how conviction becomes transferable.

The strongest founders don’t just build the product. They build the belief system around it.

The fundraising lesson: every round gets easier only after the company gets more real

David said something I think every founder should hear:

Every fundraise is different.

That sounds basic, but it matters because founders often assume fundraising difficulty is a straight reflection of founder quality. It isn’t.

Sometimes the round is hard because the market is skeptical.
Sometimes because the category lacks clear winners.
Sometimes because your story is still ahead of what the metrics can defend.

For Juicebox, the early rounds were hard. They spoke to over 100 investors. The recurring objection was clear: we haven’t seen many big outcomes in recruiting tech, so why will this be different?

That’s a hard objection to solve because it isn’t a product objection. It’s a market belief objection.

And market belief objections usually don’t get solved with better decks. They get solved with evidence.

By the time Juicebox got to the Series A, they had enough traction to make the category feel more real. The company had de-risked more of the business. The story no longer had to do all the work on its own.

That’s such an important lesson.

Early rounds are often about whether an investor believes your future could exist.
Later rounds are more about whether you’ve proven enough of it already exists.

Founder psychology: you almost have to be unreasonable

David had a great line about founders needing an unreasonable belief in themselves.

I think he’s right.

Objectively, being a founder is often a bad trade on paper. The risk is high. The odds are poor. The stress is real. The expected outcome is uncertain. So if you assess it with perfectly balanced logic, many people would never do it.

The founder’s edge is not that they ignore reality. It’s that they internally price their own capacity to figure things out much higher than outsiders do.

That’s what allows them to move.

David described it as believing the actual risk is lower because you trust yourself to solve what comes next. That feels accurate to me. It isn’t pure optimism. It’s more like self-authored probability.

You think: yes, this is hard, but I back us to make it less hard than it looks.

And without that internal math, I’m not sure many companies ever get started.

Why Silicon Valley still matters

Toward the end of the conversation, we touched on why Silicon Valley is still so special.

David’s answer was simple and strong: if you want to build the best possible company, it helps to compete where the best companies are being built.

I completely agree.

The point isn’t geography for geography’s sake. It’s that the density of talent, ambition, best practices, networks, capital, and standards forces you upward. The bar is clearer. The pace is faster. The expectations are higher.

And that matters.

Because in weaker ecosystems, even good outcomes can be capped by weaker patterns. In the best ecosystem, you see how the best companies hire, raise, ship, position, scale, and win.

The Valley still has that concentration effect. That doesn’t mean great companies can’t be built elsewhere. Of course they can. But if you want to be in the arena where the standard is set, this is still the place.

Closing thought

If I compress the whole episode into one sentence, it’s this:

The strongest companies are built by founders who choose rising-pressure markets, stay close enough to the problem to pivot when needed, and then combine speed, signal quality, and conviction into a product customers can feel working for them.

That’s what I took from David’s story.

Not just that Juicebox is building something useful.
But that underneath it is a founder pattern I respect a lot: personal connection to the problem, willingness to reset, speed without chaos, and belief strong enough to survive a market that doesn’t immediately say yes.

That combination is hard to fake.
And when it’s real, it compounds.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Anthony Lye. Anthony is one of those rare operators who has lived through multiple generations of Silicon Valley up close: from the early workstation and networking era, through enterprise software, SaaS, cloud, and now AI. What makes his perspective valuable is not just that he has seen multiple waves. It is that he has seen what those waves do to companies that assume success is durable.

What made the conversation unusually valuable wasn’t a single tactic. It was the way geography, ambition, leadership discipline, disruption, and customer truth all tied into one continuous PMF story. You cannot separate them. The quality of leadership shapes whether a company finds PMF, whether it keeps PMF, and whether it survives the next platform shift.

Let me walk you through what stood out.

The Valley itself: why Silicon Valley keeps producing reinvention
Anthony’s story starts before product strategy. It starts with the environment.

He grew up in the UK and was frustrated by a culture that, in his view, felt more shaped by legacy than possibility. The career paths looked fixed. The penalties for failure felt social as much as financial. And for someone obsessed with computing, that creates a ceiling very quickly.

What he found in Silicon Valley was the opposite. He found a place where younger people could outperform older people because the field itself was still being invented. He found a culture that celebrated trying, not just winning. He found an ecosystem where risk was not a stain on your record but often the price of admission.

The PMF lesson here is deeper than “move to the right geography.” It is this: product-market fit is easier to pursue in environments that reward experimentation. If your culture punishes failure too harshly, people stop making bold bets. And when bold bets disappear, category creation usually disappears with them.

That is one of the enduring advantages of Silicon Valley. It is not just capital. It is not just talent. It is the social permission to be wrong on the way to being right.

The first leadership muscle: paranoia about disruption
One of Anthony’s sharpest points was that success attracts attack.

If you are successful, someone is coming after you. Always.

That sounds obvious, but very few companies really behave as if it is true. Most successful companies gradually start acting as though their current position is earned, stable, and somewhat self-sustaining. Their product becomes familiar. Their distribution becomes comfortable. Their installed base becomes a source of false security.

Anthony’s point is that this is exactly when danger begins.

He has watched company after company dominate one era only to disappear in the next because they confused current market leadership with future relevance. They stopped worrying. They stopped re-questioning their assumptions. They started defending their current model instead of searching for the next one.

The PMF lesson buried in that is simple: finding PMF does not protect you from losing it. In fact, success can make you worse at seeing what threatens it.

The companies that stay relevant are the ones that treat disruption as a permanent operating condition, not a periodic surprise.

The second leadership muscle: knowing when your hypothesis is wrong.
Anthony framed great leadership in a way I really liked. Good leaders are not just people with conviction. They are people who know when to stop believing their own story.

That is much harder than it sounds.

In Silicon Valley, you need conviction to start. You need belief to recruit. You need narrative to raise money. But if that belief hardens into attachment, it becomes dangerous. You stop seeing what the market is telling you. You stop listening to weak signals. You start interpreting every objection as resistance rather than information.

Anthony’s view is that strong leaders run the company on hypotheses, not ideology. They form a view of where the market is going, they build toward it, but they remain willing to prove themselves wrong. They adjust. They cancel. They shut things down. They move capital and energy away from ideas they wanted to work but do not.

That sounds clinical, but it is actually one of the most humanly difficult things in company-building. Because the hardest part is not launching new things. The hardest part is emotionally detaching from the things you have already poured yourself into.

The PMF lesson here is this: conviction gets you into the market, but detachment keeps you aligned with it.

Discontinuity theory: markets are not made, they are re-segmented.
Anthony described disruption through a lens he called discontinuity theory, and I think it is one of the most useful ways to think about PMF.

His argument is that markets are not usually “created” from nothing. More often, they are re-segmented. Something changes, and value moves. That change might come from regulation, a new distribution path, the emergence of standards, or a technology shift. But the important point is that the winner is usually not inventing demand from thin air. The winner is reorganizing how demand gets served.

That is a very practical PMF framework.

It forces you to ask not just “What are we building?” but “What changed that now makes a new answer possible?” It also forces you to ask the much sharper question Redpoint Ventures once asked Anthony: whose ox gets gored?

That question matters because for every winner, there is usually a loser. Revenue is not sitting untouched, waiting politely for your startup to arrive. If you are building something meaningful, you are almost always trying to take budget, workflow, time, margin, or strategic control away from an incumbent model.

That is what real disruption looks like.

Netflix, Dell, and the real mechanics of PMF shifts.
Anthony used a few examples that are worth sitting with.

Netflix did not just beat Blockbuster because it had a better brand or a nicer interface. It exploited a discontinuity. DVD technology was easier to ship than VHS, and mail-order distribution was structurally better than the retail footprint for that product. Then, before that model could calcify, Netflix disrupted itself again by moving to streaming.

That second move is the more important one. It was not just a story of innovation. It was a story of self-cannibalization. Netflix was willing to weaken the model that was currently working because it could see where the world was going.

The same logic shows up in Dell. Dell did not invent the PC. It changed the distribution model by selling direct. That sounds like a commercial choice, but it had category-shifting implications. The product stayed largely the same. The route to market changed. And that was enough to produce a new winner.

The PMF lesson here is one founders often miss: product-market fit does not always come from inventing entirely new technology. Sometimes it comes from seeing that the old product can be delivered, configured, bought, or consumed in a radically better way.

That is still a PMF breakthrough. It just happens on the distribution layer rather than the core technical layer.

Why incumbents lose: they listen to too many people who want tomorrow to feel like today.
This was one of Anthony’s strongest points.

When companies go looking for customer feedback, they often talk to too many people in the early majority, late majority, or laggard segments. Those customers are not trying to imagine a different future. They are trying to preserve a familiar present. They want tomorrow to feel like today.

That is understandable. Most people do not like change. They resist it. They defend existing workflows. They overvalue stability. And if you take too much guidance from that group, you can talk yourself out of the future.

This is where many incumbents get stuck. Their best customers tell them not to move too fast. Their biggest accounts resist the new model. Their sales force prefers the old playbook. Their product teams optimize for continuity. And by the time the shift is undeniable, someone else already owns the new ground.

The PMF lesson here is uncomfortable but important: customer centricity does not mean obedience. Listening matters, but segmenting who you listen to matters just as much.

Sometimes the future enters through edge behavior before it becomes mainstream demand.

SaaS to AI: why Anthony believes AI changes the structure, not just the interface.
Anthony made a point that I think a lot of enterprise software leaders still underestimate: AI is not just a better feature layer for SaaS. It changes what the software is supposed to do.

Traditional SaaS systems were largely passive. You logged in, reviewed data, updated fields, clicked through workflows, and then logged out. The software stored process. It created visibility. It drove accountability. But in many cases, it did not actually produce the outcome.

That is why he made the point that Salesforce never really helped a salesperson sell. It helped manage the process around selling.

AI changes that.

If software can now take on the administrative burden, generate work overnight, summarize, recommend, prioritize, reason, and act, then the value moves from tool-based interaction to outcome-based execution.

That changes the product standard. It also threatens a huge number of legacy software companies that still think about AI as additive rather than structural.

Anthony’s view is that some large SaaS companies are treating AI as evolutionary because they assume their installed base protects them. He thinks that is exactly the kind of thinking that gets companies crushed.

The PMF lesson here is clear: when the standard shifts from “system of record” to “system of action,” companies that keep optimizing the old standard may discover too late that they are no longer solving the right problem.

Software plus services: the next blur.
Another important thread in the conversation was Anthony’s belief that AI blurs the line between software and services.

For years, many enterprise software companies sold the software and left the customer to figure out outcomes through consultants, implementation partners, and service layers. The software owned the tool; the service provider owned the result.

AI puts pressure on that split.

If software can now deliver more of the labor, reasoning, execution, and adaptation that previously required humans, then the software company starts to move up the stack. It is no longer just providing a platform. It is increasingly responsible for delivering the work.

That is why Anthony talked about software as labor. And that framing matters, because labor budgets are much larger than software budgets.

The PMF implication is huge. The next great enterprise companies may not just sell software licenses more efficiently. They may capture budgets that historically belonged to service firms by contracting closer to outcomes.

If that happens, category boundaries change fast.

Customer truth: opinion exists on the inside, fact exists on the outside.
The line from Anthony that stuck with me most was this: opinion exists on the inside, and fact can only exist on the outside.

That is one of the cleanest summaries of product-market fit I have heard.

Inside the company, you have beliefs, strategy, enthusiasm, politics, roadmap logic, internal consensus. Outside the company, you have the only thing that actually matters: customer behavior.

Customers tell you the truth through adoption, retention, expansion, urgency, usage, replacement, referrals, and willingness to pay. Not through your internal meetings. Not through your deck. Not through your conviction alone.

The best leaders never confuse internal agreement with external validation.

That is why Anthony recommended Bob Goodson’s The Button That Changed the World. The small act of gathering feedback, measuring response, and observing what people actually do is often more valuable than any executive narrative.

PMF is not what you say your product does. It is what the market repeatedly confirms it does.

The real takeaway: you must keep re-earning relevance.
The thread connecting the entire conversation is this: in Silicon Valley, relevance is rented, not owned.

You can build something brilliant. You can hit timing perfectly. You can ride a major platform wave. You can dominate a category. And still, if you stop re-evaluating the world, if you stop challenging your assumptions, if you listen too much to the people defending the old model, or if you fall in love with your current success, you become vulnerable.

Great leaders do not just build companies. They build systems that can question themselves.

That is what keeps PMF alive across cycles.

Closing thought:
If I compress the entire episode into one sentence, it is this: The companies that survive Silicon Valley are not the ones that invent once, but the ones that keep re-segmenting the market before someone else does it to them.

That discipline is rare. But it is the difference between becoming a chapter in a great company’s story and becoming the company that writes the next one.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Vaibhav Nadgauda. Vaibhav is one of those rare people who has seen software from every angle: engineer, operator, founder, investor, and now CEO. He started as a programmer, scaled teams and businesses, built successful companies in Sacramento, then went into venture with Moneta Ventures, and today leads App Orchid after first backing the company as an investor.

What made the conversation unusually valuable wasn’t just his investor perspective. It was the way operator experience, venture judgment, and founder resilience all tied back to one core PMF truth:

Product-market fit is rarely obvious in real time, and the teams that win are usually the ones that can see around corners before the market catches up.

Let me walk you through what stood out.

The first lesson: stick to your knitting

One of the clearest things Vaibhav said about Moneta’s early years was simple:

“Stick to your knitting.”

Their strength was enterprise software and B2B businesses. That was the world they understood deeply. And like a lot of first-time investors, they learned that common sense only takes you so far when you move outside your lane.

I think this is one of the most underrated PMF lessons in venture and in startups.

Founders often assume intelligence is transferable across markets. Investors sometimes do the same. But PMF is deeply contextual. What works in enterprise software does not automatically translate into consumer. What works in one buyer motion doesn’t automatically transfer to another. What looks obvious from a distance can become very unclear once you’re in the weeds.

The practical lesson is that domain depth matters more than general confidence. Pattern recognition only works when the pattern is one you’ve actually lived through before.

Why experience matters more than theory

Vaibhav shared a quote from his brother that I thought was brilliant:

“Experience is the most expensive teacher because everybody else is underpaid.”

That line says a lot about how he thinks.

He made the point that seeing around corners does not really come from theory. It comes from having lived through enough cycles that you can recognize what matters before everyone else does. That applies to founders, operators, and investors.

This is important because in startup land, everyone talks about vision. But vision is often just pattern recognition in disguise.

When someone has really been through enough reps, they start to spot things earlier:

● when a market is moving

● when customer feedback is noise versus signal

● when a category is about to matter

● when a company is early versus simply wrong

That doesn’t mean they’re always right. But it does mean they’re usually working from something deeper than instinct.

The hard truth: being early and being wrong often look identical

Vaibhav said something that every founder should sit with for a minute:

“Being early and being wrong are one and the same thing.”

That’s the brutal part of building.

If the market doesn’t understand what you’re building yet, it does not reward your correctness. It treats you exactly the same way it treats someone who is genuinely off-base.

That was a big part of the App Orchid journey.

The core idea behind the company was that enterprise intelligence should not live inside siloed applications and expensive consulting layers. It should live closer to the data itself, through a semantic layer that makes enterprise information usable, connected, and actionable.

That idea is obvious now in the age of AI.

It was not obvious when they started.

So for years, they had to wrap that deeper capability in narrower use cases because the market wasn’t ready to buy the foundational story. The conviction was there, but the buying language wasn’t.

That’s such an important PMF point.

Sometimes the issue isn’t that your product is wrong. It’s that your market hasn’t developed the vocabulary to understand why it matters yet.

PMF is not a destination. It is constant translation

One of the best parts of the conversation was Vaibhav’s honesty around the startup journey.

He basically said: every startup goes through the same reality. You talk to customers, you get mixed feedback, you think you know your ICP, then you realize you don’t. You keep adjusting.

That process is not failure. That is the work.

And I think this is where a lot of founders get stuck. They assume PMF will arrive like a clean, undeniable event. In reality, it often looks more like repeated translation:

● translating your technical vision into business language

● translating customer pain into a usable roadmap

● translating early signal into a focused ICP

● translating market timing into positioning

Vaibhav made the point that nobody can do this for you. Advisors cannot do it. Investors cannot do it. You have to go through the process yourself.

You have to take in the yeses and the no’s, filter them, and make hard decisions about where to focus.

The operator mindset vs the investor mindset

Another insight I loved was how Vaibhav explained the difference between being an operator and being an investor.

Operators are trained to find problems and solve them. That is their default mode. If something is broken, they dive in.

But in venture, that instinct can actually work against you.

Because if you spend all your time trying to rescue struggling portfolio companies, you can miss the opportunity to help the winners become much bigger winners.

That mindset shift is huge.

The lesson for founders is similar. There comes a point where growth is not about obsessing over every weak point. It is about identifying what is already working and pushing more energy into it.

That is just as true in product as it is in portfolio management.

Why the team matters more than the original idea

When I asked Vaibhav about the biggest traits behind his biggest investment wins, his answer was immediate:

the team.

Not the deck. Not the original product. Not even the initial market framing.

Because as he put it, what you start with is almost never what you end with.

That is one of the cleanest definitions of startup reality.

Products evolve. GTM changes. Categories shift. The founder’s job is not to be right on day one. It is to adapt intelligently without losing the underlying thread of value.

That requires:

● foresight

● resilience

● humility

● speed of learning

● the ability to carry the team through uncertainty

The team is the asset because the team is what survives the pivots.

The Moneta insight: founder empathy compounds

What also stood out was how much Vaibhav emphasized empathy.

Because Moneta’s partners came from operating backgrounds, they naturally understand the founder journey differently. They’ve lived the pressure, the ambiguity, the tradeoffs, and the psychological weight of building.

That changes how they show up.

And I think that matters more than a lot of people realize.

There is a big difference between giving advice from theory and giving advice from scar tissue.

That is also what makes the App Orchid story so powerful.

App Orchid: when backing a founder becomes carrying the vision forward

Vaibhav’s story with App Orchid is one of the most moving founder-investor stories I’ve heard.

He first invested in the company because he believed in Krishna, the founder, and in the vision. Then Krishna passed away during the early part of Covid. Vaibhav stepped in as interim CEO, and eventually as full-time CEO, because he looked at what had been built and believed the underlying vision was still right.

That says a lot.

It says that backing founders is not just about writing a check. It is about conviction in the person, the mission, and the possibility that the company still has something important to become.

And this is where the conversation became bigger than just PMF.

Because in a lot of ways, this is what real company-building is:

● carrying belief through uncertainty

● helping a team stay together

● protecting a vision long enough for the market to catch up

That is leadership.

The AI wave changed the conversation

What really struck me is that App Orchid now finds itself in a much stronger position because the market has moved in its direction.

Before, they had to educate customers on why semantics and ontology mattered.

Now, because of AI, customers are starting from a different place. They already understand they need a semantic layer to interact meaningfully with enterprise data. They already see that fragmented systems and siloed information break AI outcomes.

So the conversation has changed from:
“Why do I need this?”

to:
“How do you solve this?”

That shift is everything.

When you no longer have to spend 80% of the call explaining why the category matters, you can spend your time proving why your solution is the best one.

That is one of the clearest signs that a market is maturing.

What App Orchid is really solving

At the heart of App Orchid is a problem that every enterprise eventually runs into:

Your data is everywhere, but your intelligence is nowhere.

Customer data sits in one system. Operational data sits in another. Financial data lives somewhere else. Consultants extract it, dashboards simplify it, leadership tries to interpret it, and none of it really feels like one connected truth.

App Orchid’s thesis is that this entire model is broken.

Instead of moving all the data into giant centralized science projects and then hoping someone can make sense of it, they build a knowledge graph and semantic layer that lets enterprises interact with their data as if it were one system.

That matters even more in an AI world.

Because if AI is going to be useful inside an enterprise, it has to understand what the data means, how the entities relate, and what the real business context is. Without that layer, the output may look intelligent, but it won’t be trustworthy.

That’s why Vaibhav called it the data problem for the AI world.

And I think he’s right.

The real takeaway

If I compress the whole episode into one sentence, it’s this:

PMF often belongs not to the team that had the idea first, but to the team that stayed alive, stayed sharp, and stayed convicted long enough for the market to catch up.

That means:

● sticking to what you know

● learning through experience

● backing teams over static ideas

● listening without losing conviction

● and understanding that sometimes the hardest part of PMF is simply surviving the period when being right looks exactly like being wrong

That’s the game.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode of Inside the Silicon Mind with Hassan Khajeh-Hosseini, Co-Founder of Infracost. Hassan has spent over 15 years deep in cloud cost management, and what stands out is his willingness to stay with a problem long enough to truly understand it. That persistence led him to a simple but powerful realization: some of the biggest opportunities in tech come from fixing problems that everyone else has learned to accept. In this case, that problem is cloud cost.

The uncomfortable truth is that engineers are effectively “shopping” for infrastructure without ever seeing the price. Cloud started simple, with predictable pricing and manageable decisions. But over time, it evolved into millions of price points across providers, layered services, and deeply complex configurations. Combine that with modern engineering workflows where thousands of engineers deploy rapidly and independently and you end up with a system where decisions that can cost millions are made inside code, without visibility at the moment those decisions are made. It’s like filling a shopping cart without prices and only finding out the total a month later. That’s not just inefficient - it’s a fundamental gap.

Most companies tried to solve this by focusing on visibility. They analyzed cloud bills, built dashboards, and showed teams where money was being spent. But by the time you’re looking at a bill, the cost has already been incurred. The code is shipped, the infrastructure is live, and changing it becomes expensive and painful. Hassan’s key insight was that the problem wasn’t visibility - it was timing. Instead of showing cost after the fact, you need to show it before deployment.

That’s what Infracost does. It sits directly inside the developer workflow, analyzing infrastructure changes before they go live and showing engineers the cost impact of their decisions. Hassan describes it as a “checkout screen for cloud.” That shift - from reactive to proactive - completely changes the dynamic. It moves cost awareness from finance into engineering, and from reporting into decision-making.

One example from the conversation makes this tangible. An engineer made what looked like a normal change - new feature, new database, additional infrastructure. Infracost flagged the impact: over $500,000 per month. The team paused, rethought the approach, and reduced the cost dramatically before anything went into production. No rework, no rollback, no firefighting. This is where the real value lies. It’s not just about saving money - it’s about preventing unnecessary costs from ever happening.

And that’s where the deeper PMF shift sits. Infracost isn’t just a tool for visibility - it’s a system that changes behavior. Traditionally, cost sat with finance while engineers focused on shipping. That created tension after the fact. But when cost becomes part of the developer workflow, engineers start making better decisions upfront. Senior engineers reinforce best practices, and over time, cost awareness becomes part of the culture. This is what makes the product sticky. It’s not just solving a problem - it’s embedding itself into how work gets done.

The market itself is crowded, with over a hundred companies offering variations of “we save you money.” That simplicity attracts competition, but it also hides how difficult the problem really is. Most players sit on top of billing data and compete on dashboards or features. Infracost took a harder path by moving upstream into code, solving a more complex technical problem. And that’s often where defensibility comes from - doing the less obvious, more difficult thing.

Distribution played a major role in reinforcing that. The product was built for engineers, with free usage and open-source elements, which allowed it to spread organically. Engineers adopted it, brought it into new companies, and over time it became part of their workflow - and even something they referenced on their CVs. That’s when you know a product has moved beyond usage into real adoption. PMF isn’t just about building something valuable; it’s about how that value spreads.

Pricing followed the same philosophy. Instead of charging a percentage of cloud spend, which can feel like a tax at scale; Infracost charges per engineer. This aligns the product with its actual users and makes adoption frictionless. It’s a reminder that pricing isn’t just a monetization decision; it’s part of the product experience itself.

Another interesting dynamic is who actually cares about the problem. There’s a common assumption that engineers don’t care about cost. In reality, senior engineers care deeply because they’ve seen inefficiency, waste, and the consequences of poor decisions. They become the internal champions, driving adoption and influencing the rest of the team. Often, the real buyer isn’t the person holding the budget - it’s the operator who feels the pain.

We also touched on the tension around AI. While there’s clear potential to layer AI into the product - automating optimizations and generating improvements - many enterprises still restrict its use due to concerns around data and reliability. Hassan’s approach is pragmatic: deliver value without AI first, then introduce it gradually as customers become comfortable. PMF isn’t just about what’s possible; it’s about what the market is ready to adopt.

One of the most valuable lessons came from a failed company Hassan built. The idea was strong - simulating real outages to test DevOps candidates in realistic scenarios. But the market didn’t support it. The best candidates are scarce and expect a high-touch experience, not rigorous testing environments. Even with a great product and a real problem, there was no product-market fit. It’s a reminder that the market itself has to cooperate.

Hassan’s definition of PMF is one I agree with deeply. It’s not a moment you reach and then keep. It’s something you constantly earn. You feel it when demand surges, when inbound grows, when systems start to strain. But markets evolve, competitors catch up, and customer expectations shift. If you stop adapting, you lose it. PMF is an ongoing process across both the product and the market.

If I compress this entire episode into one sentence, it’s this:

The best PMF opportunities aren’t about building better dashboards - they’re about moving the moment of decision closer to where the problem actually happens.

Infracost didn’t win by showing companies what they spent. It’s winning by showing engineers what they’re about to spend - before it’s too late. That shift, from reactive to proactive, is where the next generation of category leaders will be built.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

When a company becomes a $100 billion public giant, the narrative tends to sound inevitable. Visionary founders. Perfect strategy. Flawless execution.

But if you talk to the people who were actually there in the early days, the story sounds very different.

Jonathan Claybaugh was employee number thirteen at Snowflake.

When he joined, there were zero customers, zero revenue, and fewer than twenty people trying to build something that looked borderline insane: a tiny startup attempting to compete with Amazon’s Redshift.

From the outside, that bet made no sense.

From the inside, something felt electric.

And that feeling - more than spreadsheets, TAM calculations, or investor decks is often the first signal that product-market fit might be coming.

The garage is real - but the magic is the room

Snowflake’s first real office wasn’t glamorous.

It was a run-down brick building in San Mateo with ratty carpets and terrible bathrooms. A classic Silicon Valley startup space: cheap, chaotic, surrounded by coffee shops.

But the office wasn’t what made the company special.

The people did.

Jonathan describes walking into a room filled with database PhDs and engineers who were literal luminaries in their field - people who had previously built world-class systems at Oracle.

What struck him wasn’t just their intelligence.

It was their mindset.

Everyone was collaborative. Everyone took responsibility. Everyone executed. If someone gave you a task, you finished it because you didn’t want to slow the team down.

It sounds like a cliché.

But it’s a cliché because it’s true.

When a small group of exceptionally capable people share a mission and trust each other to deliver, something unusual happens.

The work becomes fun.

Even before success.

The first years are just building

What’s often forgotten about Snowflake’s story is how long it spent simply building.

When Jonathan joined, the company had a proof of concept running on AWS - a rudimentary system that could query data from S3 using EC2 compute.

That was it.

The founders had demonstrated that the architecture could work. Now the team had to turn it into a real product.

While engineers were building the platform, the sales team was doing something equally brutal: cold-calling potential customers and offering the product for free just to get someone to try it.

Three salespeople hammering the phones.

Engineers racing to build something stable enough to demo.

This is what the earliest stage actually looks like: chaos, iteration, and long hours.

Not glory.

The architectural insight that mattered

Snowflake’s breakthrough wasn’t marketing.

It was architecture.

At the time, most data warehouse systems - including Amazon Redshift - inherited a design where data and compute were tightly coupled.

If you wanted more performance, you had to add more machines and redistribute the data. Scaling was painful. Resizing clusters meant moving massive amounts of data.

Snowflake flipped the model.

Data lived in object storage (S3). Compute was separate and elastic. Queries could spin up independently without reshuffling the entire system.

This sounds obvious now.

At the time, it was radical.

And that radical difference is what allowed a 20-person startup to compete with Amazon.

Not by doing the same thing slightly better.

But by doing it fundamentally differently.

The moment you realise something is working

For the first few years, nobody knew how big Snowflake might become.

Even internally.

Jonathan says the moment things started to feel real was several years in, after the Series C round, when revenue began doing something unusual:

It kept tripling year after year.

At first you dismiss it as early momentum. Then it happens again. And again. And again.

Eventually the pattern becomes undeniable.

But even then, most employees underestimated the scale of what was coming.

Some finance people inside the company were already saying: You’re undervaluing this.

Most engineers just thought: We’re doing pretty well.

That’s the strange bubble of hypergrowth startups.

Inside the company, you’re too busy building to fully appreciate what’s happening.

The metric that mattered most

Snowflake’s growth wasn’t just about landing new customers.

The real magic was expansion revenue.

Customers who started using Snowflake didn’t just stay. They used more and more of it.

Their data volumes grew. Their workloads increased. Their teams adopted the platform across departments.

Existing customers doubled their usage - sometimes repeatedly.

When you combine:

●     Strong new customer acquisition

●     Exceptional retention

●     Explosive expansion revenue

You get a growth engine that becomes nearly unstoppable.

This is the holy grail of product-market fit.

Not just adoption.

Compounding adoption.

Pride in the craft creates product-market fit

Jonathan believes Snowflake’s success came down to something simple:

The team took pride in what they were building.

That pride drove long hours. It drove craftsmanship. It drove engineers to think years ahead - like implementing support for processor instructions that didn’t even exist yet, simply because they knew they were coming.

When people care deeply about the thing they’re building, two things tend to follow:

First, the product becomes genuinely good.

Second, the team listens to customers with humility.

Jonathan puts it bluntly: if you think you already have all the answers, you’re probably wrong.

The best companies listen closely to customers, adjust their roadmap, and evolve without abandoning their core design principles.

That’s how the growth flywheel begins.

Why Snowflake’s team looked different

One unusual aspect of Snowflake’s early team was age.

Most engineers weren’t twenty-something prodigies.

They were in their thirties and forties.

People who had already built companies. Failed. Learned. Built again.

The founders themselves were in their forties when Snowflake started.

In a valley obsessed with youth, that’s a useful reminder: experience compounds too.

Sometimes the best startup teams are not the youngest.

They’re the most seasoned.

The employee strategy few people talk about

Jonathan also highlights something rarely discussed in startup advice: how employees should evaluate opportunities.

Before joining Snowflake, he had worked at several startups that failed.

So this time he approached it differently.

He looked at three things.

First, market size. Snowflake’s initial total addressable market was estimated at $15 billion. Even capturing a small percentage could create a massive company.

Second, product differentiation. The architecture was radically different from existing data warehouses.

Third, equity math.

Jonathan literally built spreadsheets estimating dilution, potential exits, and post-tax outcomes.

If the company reached even a modest outcome, the numbers worked.

This kind of intentional thinking is rare among early employees but it’s often what separates people who benefit from startup success from those who miss it.

Not every outcome needs to be Snowflake

One of Jonathan’s most refreshing points is this:

You don’t need a $50 billion exit.

If you own a meaningful percentage of a company that exits for $100 million, that can still be life-changing.

The Silicon Valley narrative often focuses only on unicorns.

But for founders and early employees, ownership matters more than headline valuations.

Startups are a philosophical choice

When young engineers ask Jonathan whether they should join a startup, he doesn’t give them a simple answer.

He asks a different question:

What kind of life do you want?

If you want stability, predictable hours, and a comfortable career path, large companies offer that.

If you want adventure - the rush of building something new, the risk of failure, and the possibility of extraordinary outcomes - startups offer that.

Neither choice is wrong.

But startups demand something very specific:

You have to want it.

Because when things get hard - and they always do - passion is the only fuel that keeps people going.

The real meaning of luck

Jonathan describes Snowflake as luck.

But his definition of luck is familiar to anyone who studies great companies.

Luck is when preparation meets opportunity.

He had spent years working on infrastructure, networking, security, and cloud systems. He had survived multiple failed startups.

When the Snowflake opportunity appeared, he had the skills - and the instincts - to recognize it.

That’s what looked like luck.

The final lesson

Looking back, Jonathan distills the Snowflake story into a few simple ingredients:

A huge market.
A radically different idea.
A team of exceptional people.
Relentless work.
Humility toward customers.

None of these guarantees success.

But when they come together in the same room, something powerful happens.

And sometimes - just sometimes - that electric room builds a generational company.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode of Inside the Silicon Mind with Leonid Igolnik. Leonid has spent decades building and leading in enterprise software, sits at the intersection of engineering, AI, and operating reality, and he has a rare ability to talk about the future without losing the thread of what actually breaks in production.

What made this conversation valuable wasn’t a prediction about agents or a hot take about “one engineer startups.” It was something more durable:

AI is changing software by changing the abstraction layer and when abstraction shifts, the winners aren’t the people who generate more code. The winners are the people who can express intent clearly, master their domain deeply, and prove correctness in a world where models are non-deterministic.

Let me walk you through what stood out.

Agents: the most popular thing nobody can define

We started where the entire industry seems to be starting: agents.

And Leonid’s answer was honest in the best way:

Nobody really knows what agents are yet. But everyone wants one.

That’s not cynicism - it’s a signal.

When a term becomes unavoidable before it becomes precise, it usually means the market is grasping for a new interface. People feel the shift coming. The label is just lagging behind the reality.

What he did anchor on is the only part that matters right now:

 ● We’re in a period of heavy experimentation

 ● Tooling is being built everywhere

 ● And it’s not clear what will win

That ambiguity is the point. It’s also why this moment is so intoxicating if you’re building.

AI is a new abstraction layer and abstraction always creates a new PMF battlefield

Leonid framed AI the way strong engineering leaders frame every platform change:

AI isn’t just a tool. It’s a new level of abstraction.

And if you’ve been around long enough, you’ve seen this movie before:

 ● Assembly → higher-level languages

 ● Rack-and-stack → cloud

 ● Manual ops → managed infrastructure

 ● Static UI → API-first

 ● Now: code-as-output → intent-as-input

Each time abstraction rises, two things happen at once:

1. More people can build more things
   

2. New failure modes explode underneath the surface

The illusion is: “this will reduce complexity.”

The reality is: it shifts complexity into different layers - often into layers you can’t ignore anymore.

The uncomfortable truth: AI produces bugs at the same rate but now we ship more code

This was one of the most sobering lines in the entire conversation:

The machinery is producing as many bugs as humans, percentage-wise… except now we’re writing way more code.

That’s the hidden tax of AI-assisted engineering.

You get speed.
You also get volume.
And volume amplifies risk.

The market is celebrating output. But the real cost shows up later:

 ● regressions

 ● brittle edge cases

 ● subtle security holes

 ● operational fragility

 ● “it worked yesterday” incidents

This is why the “AI makes engineers 10x” narrative is incomplete.

AI can make output 10x.
But it can also make failure 10x if you don’t build guardrails.

Why testing is coming back - not as a best practice, but as survival

Leonid believes testing is going to have a resurgence, and not because engineers suddenly love writing tests.

It’s because we’re trying to do something logically inconsistent:

We’re asking for a stochastic (non-deterministic) system to produce deterministic software.

If you prompt the same model with the same request three times, you can get three different outputs.

So how do you build production-grade systems in a world where the generator is probabilistic?

His answer is simple and very practical:

You need test coverage as the safety net.

Not as dogma.
As a stabilizer.

Testing becomes the method by which you transform non-deterministic generation into deterministic behavior.

And this is where something interesting happens:

What used to be tedious becomes easier because AI can help write tests, expand coverage, and describe behavior in plain English.

Which leads to a bigger point…

“English is the new programming language” and it changes who can contribute

Leonid referenced the idea (popularized by Karpathy) that:

English is becoming the interface.

But he pushed it further into a workflow-level insight:

We can increasingly express the expected behavior of software in plain language and that creates shared understanding across roles.

Historically, the chain looked like this:

 ● PM writes requirements

 ● design translates into screens

 ● engineers implement

 ● QA validates

 ● customers discover what the product actually does

Now we can do something more powerful:

 ● write tests that define expected outcomes in plain English

 ● let product and engineering align on behavior early

 ● turn “what should happen” into executable truth

That’s not just a tooling improvement.

That’s a coordination improvement.

And PMF lives and dies on coordination.

The new engineering moat: domain mastery + intent expression

Leonid kept returning to this theme:

Mastery matters more than ever. Domain expertise matters more than ever.

Why?

Because AI amplifies skill - it doesn’t replace it.

He used an analogy I loved:

A steam hammer in the hands of a skilled craftsman is far more powerful than a regular hammer - but the craftsman still needs to understand how metal behaves.

AI works the same way:

 ● In the hands of someone who understands fundamentals, it multiplies capability

 ● In the hands of someone without fundamentals, it multiplies mistakes

This is where the market is heading:

The engineer who can “use AI” won’t win.

The engineer who can:

 ● define what matters

 ● express intent clearly

 ● reason from first principles

 ● validate outcomes

 ● operate systems safely

  …will win.

Because engineering is no longer just “writing code.”

It’s building systems that behave predictably under real-world constraints.

Hiring is changing: it’s not “can you code?” - it’s “can you be a professional with tools?”

One of the most practical parts of the episode was how Leonid thinks about interviews.

He described the “tipping point” his org hit:

At first, they tried filtering candidates using AI tools.

Then they crossed the bridge and accepted the real truth:

The tools are here to stay.

So the question becomes:

How do you evolve the interview loop to:

1. allow candidates to use modern tools
   

2. still evaluate them as actual software professionals

And there’s a second-order problem that most teams ignore:

Interviewers need retraining too.

Because evaluating someone with access to models, docs, and generated code is fundamentally different than evaluating them in a constrained environment.

This is the new bar:

 ● Can you steer the tool?

 ● Can you critique output?

 ● Can you validate correctness?

 ● Can you secure what you ship?

 ● Can you explain the tradeoffs?

That’s the definition of “seniority” now.

Why the “1–2 engineer billion-dollar startup” story breaks in the real world

I asked Leonid directly: do you see the next billion dollar startup being built by one or two engineers?

His answer: No.

And his reasoning matters for founders:

Because building a company is not only building software.

You still have to:

 ● acquire customers

 ● service customers

 ● support customers

 ● run operations

 ● manage reliability and trust

 ● build redundancy (because humans need vacations)

AI can compress effort inside functions.

It can’t delete the functions themselves.

His analogy was crisp: you can fly as a solo pilot - but United Airlines has dispatchers, meteorologists, ground crews, planners, customer support, and redundancy for a reason.

Scale forces specialization.

AI changes the yield.
Not the laws of reality.

The meta-takeaway: the great filter is happening

This might be the most important line of the entire conversation:

Expertise gets amplified. Lack of expertise can’t benefit from this.

We are entering an era where mediocrity becomes expensive - not morally, economically.

Because when tools raise the baseline, the market stops rewarding “good enough.”

That sounds harsh, but Leonid offered the counterbalance that makes it hopeful:

AI also democratizes access to knowledge and best practices.

It’s a free expert coach.
It can help anyone willing to invest in themselves level up faster than ever.

So yes - the bar rises.

But the ladder is also getting easier to climb.

Closing thought

If I compress this episode into one sentence, it’s this:

In the AI era, PMF belongs to teams who can express intent clearly, master their domain deeply, and prove correctness in a world where code is cheap but reliability is everything.

AI will keep accelerating.

But the differentiation won’t be “who can generate more.”

It will be who can build systems that behave, scale, and stay trusted when the machinery is probabilistic.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

There are moments in history you never forget.

Clark Golestani can tell you exactly where he was on Black Monday in 1987. Sitting at MIT. Lunch on the table. A friend with a Quotron machine watching the ticker collapse in real time. Sprouts flying across the table as portfolios evaporated. Phone calls to Fidelity that couldn’t get through. Students who had invested their tuition loans not returning the next semester.

It was chaos.

But Clark wasn’t just watching markets fall. He was building a startup at the time - raising capital into a market that had suddenly gone silent.

Venture funding didn’t “slow.” It disappeared.

And that experience shaped a philosophy he has carried from Oracle to Merck to private equity to venture capital - a philosophy every founder should internalize:

Never get over your skis. Manage cash tightly. Even when it’s sunny, prepare for rain.

This edition of The PMF Playbook is about cycles, discipline, bubbles, and the fundamental difference between building enterprise value and building EBITDA.

Because if you misunderstand those forces, the market will teach you the hard way.

Black Monday and the illusion of permanence

What Clark learned in 1987 wasn’t just about volatility. It was about fragility.

Capital markets feel permanent when they’re working. Funding feels abundant when it’s flowing. Valuations feel justified when they’re rising.

Until they’re not.

As he was raising his next round, venture firms that had been active simply stopped investing. Meetings continued. Interest was polite. But checks weren’t being written. For one to two years, capital was effectively frozen.

That moment crystallized something most founders only learn once: access to capital is cyclical. It is not a right. It is not permanent. And it can vanish without warning.

The founders who survive aren’t the most optimistic. They’re the most prepared.

Cash is not cowardice. It’s leverage.

Clark has spent time inside hypergrowth companies and inside conservative ones. The contrast is instructive.

At Merck, under CFO Judy Lewent, the company was famously conservative. Cash buffers were built quietly. Critics questioned why so much capital sat idle.

Then Vioxx happened.

A multibillion-dollar drug was voluntarily pulled from the market. Any other company might have collapsed under the financial shock. Merck survived because it had prepared for a rainy day long before the clouds formed.

Clark sees founders make the opposite mistake constantly. They raise late. They burn aggressively. They treat runway like a shot clock in basketball, pushing right to zero before negotiating the next round.

But time is leverage.

If you’re negotiating with six weeks of runway left, the investor has the power. If you’re negotiating with 18 months in the bank, you do.

And leverage determines terms.

Clark’s rule is simple. Never let runway fall below six months. Ideally, never below twelve. And if you can engineer 24–36 months at earlier stages, even better.

For profitable businesses, the principle remains the same. Keep twelve months of cash on hand. Negotiate credit facilities when times are strong, not when you’re desperate. Banks lend when you don’t need it. They disappear when you do.

Cash isn’t fear. Cash is optionality.

Spotting bubbles without missing revolutions

Clark also carries another scar from 1987: the importance of spotting bubbles.

He believes AI today sits squarely in a hype cycle. That doesn’t mean it isn’t transformative. It likely will be more transformative than IT, perhaps rivaling the steam engine in productivity impact.

But hype and inevitability can coexist.

The mistake isn’t believing in the transformation. The mistake is confusing the current valuation environment with the long-term value creation curve.

Clark distinguishes between evolution and bubble this way: humans overestimate what will happen in one year and underestimate what will happen in ten.

AI will change everything. But we are still in the “brick phone” phase - the large suitcase-era mobile phone, not the iPhone.

The risk today isn’t that AI won’t matter. It’s that people will overpay for proximity to it, underprepare for volatility, and underestimate the time required for infrastructure - compute, energy, quantum - to truly support the most extreme visions.

In other words: believe in the future. Price the present carefully.

Venture vs. private equity: two languages, two religions

Clark operates in both worlds - private equity and venture - and he describes the difference in almost theological terms.

In private equity LP meetings, one word dominates every conversation: EBITDA.

Profitability. Covenants. Leverage ratios. Debt coverage. Margin expansion. Everything flows through the lens of sustainable cash generation and risk management.

The objective is disciplined value creation with managed risk. A 3–5x return is excellent.

In venture LP meetings, the vocabulary shifts entirely.

Enterprise value. Market expansion. Category dominance. Speed. Optionality. Risk-taking. A 3x return is insufficient. The expectation is asymmetric upside - Snowflake, Google, generational outcomes.

Private equity optimizes for durability and EBITDA growth.

Venture optimizes for enterprise value and outsized market expansion.

Neither is superior. They serve different mandates and different LP expectations. Wealth preservation and wealth multiplication require different strategies.

But founders must understand which game they are playing.

If you are venture-backed, you are being funded for scale and enterprise value expansion, not modest profitability.

If you are private equity-backed, profitability discipline is not optional. It is foundational.

Confuse the two, and you disappoint your investors.

Growth can kill you too

One of Clark’s most counterintuitive lessons came from Oracle.

When he joined, Oracle was growing at over 100% per year. Explosive growth. Market enthusiasm. Hyper-expansion.

And yet, that speed nearly killed the company.

Hypergrowth introduces its own risks. Quality slips. Systems lag. Leadership gaps widen. The very momentum that looks like success can become destabilizing.

Clark tells a story about Snowflake’s early days under Bob Muglia. As CIO at Merck, Clark repeatedly tried to become an early Snowflake customer.

Bob said no.

For nearly three years.

The product wasn’t ready for heavily regulated industries like biopharma. Rather than chase revenue prematurely, Snowflake sequenced its market entry carefully.

Paradoxically, every “no” built more credibility.

When Snowflake finally entered biopharma, it did so with trust.

The lesson is subtle but powerful: saying no can increase long-term enterprise value.

Chasing every dollar creates internal chaos and weakens your mission. Discipline creates trust.

The K2 thesis: access bends the curve

Clark’s venture fund, K2 Access, was built around a specific hypothesis: access to markets and decision-makers accelerates product-market fit.

Startups don’t just need capital. They need high-quality customer conversations, feedback loops, and credibility.

If you can compress the time to product-market fit and connect founders to buyers early, you bend the curve of success.

Early data from K2 Access is striking. A significant percentage of companies passing through its ecosystem have reached unicorn status. That’s not accidental. It’s the byproduct of curated exposure and disciplined selection.

The broader insight is this: venture value-add is not just capital. It’s network density, signal quality, and acceleration of learning.

The hard thing remains the hard thing

When asked for a book recommendation, Clark echoes what many elite founders and investors have said before: The Hard Thing About Hard Things.

Because eventually, every founder faces hard decisions.

Replacing executives. Letting early employees go. Pivoting strategy. Raising under pressure. Cutting costs. Walking away from opportunities.

Larry Ellison, Clark notes, repeatedly restructured leadership at Oracle at critical inflection points. Growth requires evolution. Evolution requires uncomfortable decisions.

If you can’t make them, you must find someone who can.

The throughline

From Black Monday to Oracle to Merck to private equity to venture, Clark’s worldview has a consistent thread:

Cycles are inevitable.
Bubbles form.
Growth can destabilize.
Capital disappears.
Leverage shifts quickly.

The founders who endure are not the most optimistic. They are the most disciplined.

They don’t get over their skis.
They manage cash tightly.
Even when it’s sunny, they prepare for rain.

Because when the storm comes - and it always does - preparation isn’t defensive.

It’s survival.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Joseph Doyle, Director of Recruiting at Harness. Joe has built teams through multiple market cycles - pre-AI hiring booms, post-pandemic resets, and now the full acceleration of AI across every function in tech.

What made this conversation unusually valuable wasn’t a single recruiting tactic. It was the way AI, hiring philosophy, interview discipline, and emotional intelligence all tied into one continuous PMF story.

Because here’s the truth most founders underestimate:

The quality of your recruiting system shapes the quality of your people.
The quality of your people shapes the quality of your product.
And the quality of your product determines whether you ever reach real PMF.

You can’t separate them.

Let me walk you through what stood out.

The AI lens: efficiency is the surface, leverage is the shift

Joe was honest about where AI is today inside recruiting.

On the surface, the gains are obvious:

● Job descriptions now take minutes
● Outreach writing is faster
● Framework thinking is easier
● Scheduling and sourcing are increasingly automated

The low-hanging fruit is real.

But the deeper shift isn’t speed - it’s reallocation.

If recruiters spend less time manually cranking administrative flywheels, they gain more time to:

● deepen candidate relationships
● understand timing and motivation
● coach hiring managers
● improve candidate experience
● refine role definition

That changes the nature of the function.

The PMF lesson buried in that shift is simple:

When AI removes the busywork, it exposes whether your team was strategic or mechanical.

Recruiting becomes less about process management and more about judgment.

And in competitive markets, judgment compounds.

The capacity shift: recruiters will carry more weight

Historically, recruiting scaled linearly.
More headcount plan → more recruiters.

AI changes that math.

Joe pointed out something subtle but important: recruiter capacity is likely to expand. Each recruiter may be responsible for more hires, more process oversight, and more stakeholder enablement.

That forces a skill upgrade.

The recruiter of the future becomes:

● systems-oriented
● process-design focused
● data-aware
● internally consultative
● externally relationship-driven

In other words, recruiting shifts closer to product thinking.

The PMF takeaway:

If recruiting becomes more leveraged, the cost of a weak recruiting function increases.

The companies that treat recruiting as a strategic lever will build stronger teams and stronger teams win markets.

The talent war in the AI era: clarity beats compensation

We discussed the obvious reality: AI talent is aggressively pursued. Some engineers are fielding offers in the millions.

Most companies can’t compete on raw compensation.

So what’s left?

Joe highlighted a key shift: candidates are asking deeper questions.

Not just:

● “What’s the comp?”
● “What’s the role?”

But:

● “How serious is your AI strategy?”
● “Is leadership actually committed?”
● “How big will this team be?”
● “What’s your trajectory in this space?”

In a noisy market, narrative clarity becomes a competitive advantage.

The PMF lesson:

When markets get competitive, talent doesn’t optimize for salary alone.
They optimize for trajectory.

If your strategy is fuzzy, great candidates will sense it immediately.

Growth mindset in the AI era: loop speed matters

Joe described the growth mindset as a “mind in motion.”

In the AI era, that motion accelerates.

Engineers can:

● close learning loops faster
● iterate faster
● prototype faster
● debug faster
● test hypotheses faster

That means a growth mindset is no longer an abstract curiosity.

It’s curiosity + iteration speed.

And that changes hiring.

Joe emphasized the importance of first principles thinking over narrow tool matching. Tools evolve quickly. Foundations endure.

The PMF implication:

When platforms shift, hiring for tools becomes obsolete.
Hiring for reasoning and adaptability becomes essential.

Recruiting will evolve but it won’t disappear

We addressed the existential question: will recruiters be replaced?

Joe’s view was pragmatic.

Recruiting will exist but its definition will change.

Automation will handle:

● coordination
● early filtering
● documentation
● parts of sourcing

Humans will still handle:

● reading nuance
● managing expectations
● coaching hiring managers
● navigating sensitive conversations
● building trust

Because hiring is not purely mechanical.

It’s emotional.

And that brings us to the next lens.

EQ: the differentiator AI can’t easily replicate

Joe made an important point: you can remove friction from a process without removing humanity.

In fact, you should.

Automate the annoying parts.
But surgically inject EQ where it matters.

Candidates remember:

● how they were treated
● how transparent the company was
● how fast decisions were made
● whether the process felt respectful

In commoditized talent markets, experience is differentiation.

The PMF lesson:

As automation increases, genuine human connection becomes a competitive edge.

Interview design: hire for potential, not proof

This was one of the strongest themes of the conversation.

Joe referenced a common mistake: companies say they hire for potential, but design interview loops to hire for proof.

Proof-based hiring sounds like:

● “Have you done this exact thing before?”
● “Solve this exact problem under pressure.”
● “Match the pattern we recognize.”

Potential-based hiring looks at:

● learning speed
● reasoning ability
● curiosity
● adaptability
● motivation

And here’s the uncomfortable reality Joe surfaced:

When you hire only for proof, you often attract people already ready to leave the level you’re hiring for.

When you hire for potential, you attract people who can grow inside your company.

That matters enormously for PMF.

Because early PMF requires elasticity.
Scaling PMF requires internal growth.

Rigid hiring creates rigid teams.

Process discipline: plan before you recruit

Joe was clear on something operational but powerful.

Too many companies start recruiting before defining:

● what good looks like
● who owns each decision
● how the loop is structured
● what signals matter
● what timeline is acceptable

The result? Delays, confusion, lost candidates.

Strong recruiting loops behave like projects:

● clear criteria
● clear stages
● clear ownership
● fast feedback cycles

The PMF lesson:

Candidate experience is not perks and swag.
It’s operational discipline.

And discipline accelerates momentum.

Take-homes and over-interviewing: friction filters the wrong way

Joe’s view on take-homes was nuanced.

Short, time-boxed, experiential exercises that replace heavier loops?
Potentially useful.

Long, unpaid projects as gatekeepers?
A great way to lose top-tier talent.

He also raised an important philosophical question:

If your process begins with heavy friction, who are you selecting for?

Often, desperation not excellence.

And that misalignment shows up later.

The PMF lesson:

Your interview process is a filter.
Design it intentionally.

The broader shift: more builders, more companies, more noise

Joe also touched on something bigger.

AI lowers the barrier to building.
Citizen developers can create things that previously required full engineering teams.

This may lead to:

● more small companies
● more experimentation
● more niche products
● faster iteration cycles

In that world, speed increases and advantage windows shrink.

Which means hiring becomes even more important.

Because if building is easier, differentiation shifts to:

● insight
● team quality
● execution discipline

The real takeaway: PMF is downstream of hiring quality

If I compress the episode into one sentence, it’s this:

AI will compress recruiting workflows, but the companies that win will be the ones that hire for potential, design disciplined processes, and preserve human connection inside an increasingly automated world.

You cannot separate PMF from people.

You cannot separate people from process.

And you cannot separate process from leadership intent.

Great products are built by great teams.
Great teams are built through deliberate hiring systems.
And deliberate systems don’t happen by accident.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

Some conversations don’t just give you “insight” - they recalibrate how you think.

That’s what happened when I sat down with Arvind Sodhani on Inside the Silicon Mind.

Arvind has lived through multiple waves of innovation - semiconductors, PCs, the internet, cloud, and now AI - but what stood out to me wasn’t the breadth of his experience. It was the discipline of his thinking. He doesn’t romanticize trends. He doesn’t inflate narratives. He reduces everything to first principles: risk, incentives, human behavior, timing, and revenue.

We talked about what most investors don’t say out loud:

● Risk is often just a polite word for “valuation.”

● Founder conflict kills companies faster than bad technology.

● AI is very real… but the economics are still hazy at today’s scale.

Here’s the episode, distilled into a playbook you can actually use.

“Risk is a function of what valuation you’re getting into an investment at.”
 — Arvind Sodhani

1) Risk Isn’t Abstract. It’s Math + Timing.

Arvind’s framing is refreshingly unsexy:

● All investing has risk.

● Risk doesn’t go away - it just changes shape.

● And at a high level, risk rises when valuations get aggressive (which is where we are in many markets today).

But the part that hit hardest was his breakdown of innovation risk:

Two types of startups:

A) Displacement startups
You’re “eating someone else’s lunch.” There’s an existing budget and an existing market. Risk is lower because customers already spend money there.

B) Market-creation startups
You’re building something nobody has bought before. Risk is higher because now you’re responsible for creating belief + behavior + budget.

And history is littered with “great products” that died because timing was off - not because the product was bad.

2) The Founder Risk Framework (That Most People Miss)

When Arvind assesses founders, he’s not looking for charisma. He’s looking for signals that the founder can survive the reality distortion field of startup life.

Here’s what he watches for:

1. Belief + passion

Not performative. Real conviction. If the founder isn’t deeply bought in, that’s an immediate red flag.

2. Founder work ethic

Arvind calls it “founder hours.” The willingness to grind relentlessly when there’s no momentum yet.

3. Persuasiveness

A founder must sell: to customers, candidates, partners, investors. And customers are the hardest to convince - especially when you’re asking them to switch.

4. Ability to build a team

A startup doesn’t scale on founder energy alone. Arvind wants to know: can they surround themselves with people better than them?

5. Co-founder dynamics (the silent killer)

This was one of his strongest points:

“Nothing destroys a company faster than when two founders clash.”

He looks for whether co-founders are complementary or competitive because conflict doesn’t just slow execution… it fractures it.

6. Coachability (without losing conviction)

This is the nuance: great founders are stubborn about the mission, flexible about the path.

Arvind’s belief:

“Virtually no company has taken their original product and grown huge on that original product.”

If a founder is closed to feedback, they don’t pivot - they rationalize. And denial can be both a superpower and a liability.

3) Intel Capital Was Built to Create a Market (Not Hedge One)

The origin story is a lesson in strategic clarity.

Intel had a microprocessor that could compute - but early on, there weren’t enough meaningful applications to make it indispensable to everyday users.

So Intel started investing in companies that would make the PC more valuable:

● applications

● graphics

● memory

● adjacent categories that increased demand for compute

In Arvind’s words: they needed “fellow travelers” - companies whose success was tied to the microprocessor’s destiny.

And importantly: The VC ecosystem wasn’t mature enough in the late 80s to naturally fund those categories at scale - so Intel stepped in.

Also: Arvind attributes a lot of this to Andy Grove’s support. Without Grove backing the concept internally, Intel Capital likely doesn’t exist.

4) The Cloud Warning Signal (And Why Corporations Miss What’s Obvious)

This section is a masterclass in how incumbents get blindsided even when the data is right in front of them.

Intel Capital had hundreds of startups in its portfolio and they started noticing a clear pattern:

● Everyone was using AWS

● Companies weren’t building in-house data centers anymore

● Intel had huge enterprise server dominance… but no meaningful cloud presence

So Arvind tried to raise the alarm internally.

The response he got?

“How big are these companies?”
 “They’re all startups.”
 “So… who cares?”

This is the innovator’s dilemma in plain English:
When you’re making historic margins in a core business, small signals look irrelevant until they become the new default.

Arvind’s broader lesson: Intel Capital created early warning signals that helped Intel see the cloud shift sooner than it otherwise might have.

5) AI Is Real. The Question Is: Who Pays?

Arvind is emphatic that AI is not hype:

● it’s displacing manual work

● it’s now displacing “brain-intensive effort”

● it’s reducing the labor of programming

● and it will unlock new products that weren’t possible before

But his concern isn’t capability.

It’s economics.

“If you add up the total investment in AI… it’s hundreds of billions of dollars annually.”
 “It’s still unclear where the additional revenue stream is going to come from.”

His point is subtle but important: Earlier waves (like the internet) often displaced existing revenue (travel agents → online booking). The money was already in the system.

With AI, the investment scale is so large it’s stressing even the biggest corporate balance sheets and it’s not yet obvious that revenue expands fast enough to justify the annual burn.

His prediction?

● AI will create entirely new business models (like the internet did)

● but not everyone will monetize successfully

● and some players will lose the race despite spending heavily

6) The Next Moat: Enterprise Inference Models + Data Monetization

Arvind believes the biggest enterprise unlock comes when companies build and run their own inference models - not just experiment with AI on the surface.

Why?

Because every enterprise is sitting on massive amounts of data they haven’t been able to fully monetize - not because it lacks value, but because it’s hard to interpret and activate.

Inference models (in his view) will enable:

● better customer experiences

● faster product improvements

● lower cost to serve

● entirely new offerings pulled from dormant data

His simple closing line here felt like the real playbook:

“Cheaper, faster, better wins.”

The Takeaways

Here’s Arvind’s PMF-and-investing lens, distilled:

1.  Risk is deeply tied to valuation.

2. Displacement is easier than market creation.

3. Founder dynamics can kill companies faster than product flaws.

4. No startup wins with the original product - adaptability is mandatory.

5. AI is real - but the revenue math at today’s scale is still unresolved.

6. The next enterprise wave will be inference + data monetization.

7. Creative destruction explains almost everything.

Final Thought

If there was one theme that kept resurfacing, it was this:

The future is rarely hidden. It’s usually visible first in the places powerful incumbents don’t take seriously.

And the people who win aren’t the loudest. They’re the ones who recognize patterns early… and have the conviction to act before the story feels obvious.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This one is from my Inside the Silicon Mind episode with Justin Fitzhugh.

There’s a moment Justin describes that perfectly captures the emotional whiplash of building in Silicon Valley.

At Snowflake’s IPO in September 2020 - the largest software IPO in history - he’s sitting at home on Zoom, mid-pandemic, stock ripping, years of pressure finally released. Excitement. Disbelief. Relief. Gratitude. A once-in-a-lifetime outcome.

And yet, that moment only makes sense when you understand the one that came before it.

Because just a few years earlier, Justin was deep inside a very different story at Instart Logic - a company with world-class talent, marquee customers, massive valuations, and all the surface-level signals of success… that ultimately didn’t make it.

This edition of The PMF Playbook isn’t about celebrating the peak. It’s about understanding the difference between looking like you have product-market fit and actually having it and why Snowflake crossed that chasm while so many others don’t.

When valuation runs ahead of reality

Justin is careful not to rewrite history at Instart as a simple failure. In many ways, the company did a lot right. The idea was strong. The execution was strong. The team scaled. Customers trusted them with mission-critical infrastructure where seconds of downtime meant millions of dollars lost.

But the subtle mistake - and the dangerous one - was letting valuation become the leading indicator instead of the trailing one.

At a certain point, funding rounds and paper valuations started to create a false sense of security. The company looked like a winner. It was talked about as a top portfolio asset. And when that happens, it becomes easier - almost unconsciously - to let momentum replace scrutiny.

The hardest part, Justin admits, is that when you’re inside it, you want to believe. You’ve invested years of your life. Your identity, your future, your relationships are tied to the company. So when warning signs appear - delayed rounds, valuations stalling, markets shifting - it’s incredibly difficult to look at them objectively.

Outsiders can see the curve bending long before insiders can.

The lesson here is uncomfortable but fundamental: valuation is not validation. In fact, high valuations often increase pressure before fundamentals are ready to support it. Expectations rise faster than revenue. Growth targets accelerate. Optionality disappears.

Instart didn’t fail because the team wasn’t good enough. It failed because the market commoditized faster than the company adapted and because course correction came too late.

Product-market fit is not a milestone - it’s a discipline

One of the most important ideas Justin returns to again and again is that product-market fit isn’t something you “achieve” and move on from.

At Instart, product direction was set and then left largely untouched for too long. Engineering, product, and customers weren’t always tightly coupled. Engineers didn’t consistently hear feedback directly. Market shifts - particularly rapid price compression in the CDN space - weren’t acted on aggressively enough.

The takeaway is sharp: PMF decays.

Markets move. Customers evolve. Commoditization creeps in. If you’re not constantly re-validating your assumptions - monthly, not annually - you’re slowly drifting out of alignment while convincing yourself everything is fine.

Justin’s advice to founders is almost deceptively simple: stay brutally close to customers, let engineers talk to them directly, and be willing to kill sacred cows. Sunk cost bias is one of the most expensive habits in startups.

Sometimes the bravest move isn’t pushing harder - it’s stopping, resetting, and pivoting before the window closes.

The Snowflake difference: changing the rules of the game

When Justin joined Snowflake, the contrast was immediate.

Within two sentences, the founders could articulate exactly why Snowflake was different: they disaggregated compute and storage and built natively for the cloud. That clarity wasn’t marketing polish — it was architectural conviction.

This is a defining PMF insight: category-defining companies don’t just compete better, they change the rules.

Snowflake didn’t try to out-Oracle Oracle. It redefined how data platforms should work. Storage became cheap and ubiquitous. Compute became elastic and value-aligned. Complexity disappeared for the customer.

And the result was something Justin had never seen before: frictionless sales. Sales teams didn’t “sell” - customers pulled the product in. Demand outpaced the organization’s ability to scale.

That’s real product-market fit. When growth becomes operationally terrifying rather than commercially uncertain.

Scaling PMF is harder than finding it

What looks effortless from the outside was anything but inside Snowflake.

Justin’s role - spanning cloud engineering, product security, observability, and release - existed to solve a single problem: how do you keep up when demand compounds faster than your systems?

Snowflake had to repeatedly reinvent how it operated. DevOps became software engineering. Humans managing infrastructure became software managing infrastructure. Entire cultural shifts were required - new skills, new hires, new ways of thinking.

This is a subtle but critical PMF lesson: success creates its own failure modes. The systems that get you to $100M ARR will break at $500M. The org design that worked at one scale becomes a bottleneck at the next.

Snowflake survived these transitions because leadership was willing to make painful step-function changes early - even when things looked good.

Leadership that builds trust, not theater

Justin’s reflections on leadership - particularly Frank Slootman’s - are striking in their consistency.

What made Snowflake work wasn’t charisma or hype. It was directness.

Frank was explicit about what was working and what wasn’t. He didn’t overpromise to customers. He didn’t sugarcoat internally. If something failed, he said so. If the company needed to pivot, he explained why.

That kind of honesty builds trust - with employees, with customers, and with the market.

And trust compounds.

The same principle shows up in Justin’s own leadership philosophy: clarity beats comfort. Teams don’t need protection from hard truths. They need alignment around reality.

The deeper lesson: false peaks versus escape velocity

Looking at Justin’s journey - Instart to Snowflake - the contrast becomes clear.

Instart hit a false peak: high valuation, strong narrative, but weakening fundamentals underneath.

Snowflake hit escape velocity: undeniable pull from the market, structural advantage, and systems built to scale into it.

Both moments felt exciting in real time. Only one endured.

For founders, this is the heart of PMF: don’t confuse momentum with inevitability. Don’t let valuation outrun learning. And don’t assume fit is permanent just because it once existed.

The market never stops moving. The job is to move with it - faster, clearer, and more honestly than everyone else.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my podcast episode with Gidi Cohen, Founder and CEO at Bonfy.ai.

Success isn’t final. Failure isn’t fatal. What matters is the ability to keep moving when the path is unclear.

That’s where Gidi Cohen starts. Not with product features, not with funding, not with pitch decks. With a mindset: chase the problems that don’t look solvable yet, and earn your edge by decomposing the impossible into the doable.

The real AI story inside companies isn’t just speed. It’s scale without supervision. The volume of generated content is rising, humans are leaving the loop, and the cost of being wrong is climbing fast. The future isn’t a single catastrophic breach. It’s a thousand tiny “harmless” moments that quietly erode customer trust until the brand is bleeding and nobody can explain where it started.

That’s the battlefield Bonfy is stepping into.

The real problem isn’t data. It’s judgment.

For two decades, “data security” has mostly meant pattern matching. Find nine digits. Flag the social security number. Detect a keyword. Trigger an alert. It feels like control, but it’s often theatre.

Gidi’s argument is blunt: the industry optimized for what was easy to build, not what businesses actually needed. The thing that matters most isn’t whether sensitive data exists in a message. The thing that matters is whether that message is appropriate in context.

A social security number sent from a customer to a service provider could be legitimate. The same number forwarded externally could be a disaster. If your system can’t tell the difference, you don’t have a security program - you have noise.

And noise has consequences. Every enterprise is drowning in tools, alerts, dashboards, and workflows stitched together under pressure. Complexity keeps rising, headcount never catches up, and teams lose the forest for the trees. In that environment, false positives don’t just annoy people. They train the organization to ignore the machine.

But Gidi’s more controversial point is the one most leaders miss: false negatives are the real crisis. If the majority of incidents aren’t even detectable by the existing approach, then the market doesn’t just need incremental improvement. It needs a reset.

Why now: GenAI turns a weakness into an existential risk

Before GenAI, bad tools were tolerated because humans were still doing most of the thinking. Yes, mistakes happened. Yes, someone occasionally sent the wrong attachment to the wrong person. But humans were still the gate.

Now, we’re moving into a world where content is produced at industrial speed, distributed across channels, and often shipped with minimal review. That changes what “risk” means.

Gidi’s investment bank example makes it real. Imagine analysts generating decks, summaries, filings, and communications through copilots and chat interfaces. What data went in? What was learned? Who validated the outputs? Who checked for hallucinations? And even if the output looks polished, what happens when it’s wrong, or when it accidentally merges client data across deals?

The failure mode isn’t theoretical. It’s already happening in quieter forms: the wrong portfolio landing in the wrong inbox, the wrong link permissions, the wrong shared repository. GenAI doesn’t introduce the first mistake. It multiplies the number of mistakes you can make per day while reducing the number of humans available to catch them.

That’s why Bonfy isn’t positioning as “a better DLP.” It’s positioning around something more fundamental: replacing pattern matching with context-based decision-making, and doing it in a way that can keep up with the volume that AI creates.

The product thesis: Adaptive Content Security

Bonfy calls it ACS: Adaptive Content Security.

The claim isn’t “we detect more patterns.” The claim is: we can understand content inside the business context of the organization and make better decisions about what’s risky and what isn’t.

Context means asking questions older tools don’t know how to ask:

Who is sending this?
Who is receiving it?
What relationship exists between them?
What rights do they have to the information?
What customer, deal, project, or system is the content tied to?
What channel is being used, and how does that change risk?

This is the shift from “what’s in the message” to “what does this mean in this moment.”

And once you accept that idea, a second insight becomes obvious: content risk doesn’t live in one channel. It moves through email, chat, file shares, SaaS apps, copy/paste workflows, and now GenAI prompts. Solving one channel doesn’t solve the business problem. That’s why the architecture has to be multi-channel by design, not bolted-on after the fact.

The PMF lesson: customers don’t design products but they expose truth

Gidi has spoken with an enormous number of CISOs, and his posture is healthy: listen aggressively, but don’t outsource judgment.

Customers can articulate pain, priorities, and constraints. They can validate whether your approach resonates. They can also surprise you and force you to update your roadmap. But if you ask customers to invent the product, you’ll build a better version of yesterday’s tool and lose to someone who rewrites the category.

This is a sharp PMF principle: learn the problem from customers, but build the solution from first principles.

That’s especially true in emerging markets, where customers are still asking for improvements to old systems because they haven’t yet internalized what the new world demands. PMF isn’t about giving buyers what they request. It’s about giving them what they’ll eventually realize they needed.

The founder operating system: optimism and paranoia aren’t opposites

One of the most useful takeaways from this conversation has nothing to do with security.

Gidi describes a daily cycle: waking up early with a list of what could go wrong, then converting that stress into a plan by the time the workday begins. That’s the founder pattern in its cleanest form.

Paranoia keeps you alive. Optimism keeps you moving. The trick is not choosing one. The trick is turning fear into execution.

It’s also why his Churchill framework lands so well. Leadership isn’t just seeing the future. It’s translating the future into actions that happen today, and rallying people around those actions long enough for momentum to form.

Where this is going

The market is about to split into two camps.

One camp will try to patch old approaches onto new GenAI workflows and call it “AI security.” It will mostly be feature marketing.

The other camp will treat the problem as what it really is: the need for machine-level judgment at machine-level scale, because humans cannot remain at the gate when content generation becomes exponential.

If Gidi is right, the winners won’t be the vendors who detect the most patterns. They’ll be the ones who help organizations maintain trust when every process becomes AI-assisted, every employee becomes a content producer, and every mistake moves faster than your ability to respond.

In the enterprise AI era, context isn’t a feature. It’s the moat.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my podcast episode with Varun Badhwar, Founder and CEO at Endor Labs.

There’s a moment in my conversation with Varun Badhwar that perfectly captures what it takes to build something that lasts.

He says: if you can share an idea with someone in five minutes over a beer and they go build a better company than you… that’s on you. Not because ideas don’t matter, but because execution is the only thing that turns an idea into a business.

And in today’s market where AI is compressing time, copying is easy, and narratives travel faster than products; execution isn’t just the differentiator. It’s the entire game.

Varun has lived this from every angle. He’s built and sold companies. He’s sat on both sides of acquisitions. He’s scaled a cloud security business from zero to hundreds of millions. And now, at Endor Labs, he’s building in the middle of one of the biggest platform shifts we’ve ever seen: software moving from “written” to “assembled.”

This edition is about what he’s learned - what founders misunderstand about exits, why most acquisitions fail, what integrity looks like when money is on the table, and why AI is creating a new category of risk that most teams aren’t ready for.

Build to be great. Exits are a consequence, not a strategy.

A lot of founders quietly build with an outcome in mind: “We’ll get acquired in two years.”

Varun doesn’t just disagree with that mindset - he thinks it creates weak companies.

Not because acquisition is a bad outcome. But because you can’t engineer it.

You can be doing everything “right” and still miss the window because the market changes, the buyer changes priorities, the category shifts, or the acquirer decides to build instead. If your strategy is built around being bought, you start cutting corners without realizing it. Product decisions are made for the demo, not for scale. Go-to-market becomes a sales story, not a repeatable engine. The company becomes optimized for a moment instead of a decade.

Varun’s alternative is simpler and harder: build the greatest company you can in your category. Build a brand that people trust. Build the kind of culture that attracts the best people. Build operationally in a way that scales without breaking. When you do that, you don’t chase outcomes, you create leverage.

And when you have leverage, you get options.

That’s the part founders forget: the goal isn’t to “get acquired.” The goal is to become the kind of company that can choose what happens next.

The real M&A question isn’t price. It’s what happens after.

Varun says something that should reshape how every founder thinks about acquisition conversations: nine out of ten acquired companies die within two to three years.

They don’t die because the product suddenly becomes bad. They die because the operating system changes.

A startup with 150 people gets absorbed into a company with 10,000. The five-person marketing team becomes a tiny part of a thousand-person function. Engineering gets pulled into roadmaps they didn’t write. The pace changes. The spirit dissolves. The people who were building with urgency find themselves living inside process.

So Varun asks a different question: what are my odds of success post-acquisition?

With RedLock, what made the deal work wasn’t the number. It was the structure. Prisma Cloud stayed intact. The business kept its own operating rhythm. It wasn’t “integrated into a machine.” It was allowed to function like a company with autonomy, accountability, and speed while benefiting from the scale, brand, and resources of Palo Alto Networks.

That’s the difference between a deal that becomes a footnote and a deal that becomes a platform.

If you’re a founder evaluating acquisition interest, Varun’s framing is a good north star: don’t negotiate like you’re selling a product - negotiate like you’re protecting a mission.

Your team is your legacy. And your reputation is your next company.

The most powerful part of this episode wasn’t tactical. It was moral.

Varun talks about founders who exit in a way that leaves employees behind - people who built the foundation of the company but don’t participate in the outcome. He’s blunt about it: it’s unethical.

And beyond ethics, it’s also strategically foolish.

Silicon Valley doesn’t forget those stories. Your next company is built on your last reputation. The engineers you’ll want later, the executives you’ll need later, the investors you’ll rely on later - they all remember how you behaved when you had leverage and when you didn’t.

Varun contrasts that with the messages he still receives years after RedLock - employees telling him they bought homes, created security for their families, and changed their lives because the acquisition worked for the whole team.

That’s the kind of detail people never put in a press release but it’s what defines leadership.

In a world obsessed with valuation, this is the quieter truth: integrity compounds.

AI is rewriting software. Not by killing developers, but by multiplying them.

Varun’s phrase is one I keep coming back to: software development is becoming software assembly.

Most modern codebases aren’t written from scratch. They’re stitched together - dependencies, open-source packages, snippets from the internet, internal libraries, and now, AI-generated code. And that shift changes everything, because it increases the speed of creation without automatically increasing the quality or safety of what’s created.

Varun doesn’t think developers disappear. He thinks the number of people producing code explodes.

Your marketing team writes code. Your ops team writes code. People who would never have learned a programming language can now “build” by prompting. That’s not a distant future - it’s already happening.

But the hidden cost is that velocity creates risk. AI has learned from the entire internet, which means it learned the good patterns and the bad ones. And without guardrails, code becomes easier to generate than to trust.

Varun shares a reality that should make every founder pause: a majority of AI-produced code can be insecure by default.

So the bottleneck shifts. The problem is no longer “can we write code fast enough?” The problem becomes “can we ship software that’s safe enough to rely on?”

Endor Labs and the supply chain era: why this problem gets bigger every year

Varun started Endor Labs before ChatGPT, anchored on a thesis: open-source software and third-party dependencies would become one of the biggest security fault lines in modern engineering.

Then Log4j happened, and the world saw what that thesis looked like in practice.

Now AI amplifies the same issue: more code, more dependencies, more unknown provenance, more risk, shipped faster than teams can review.

The old approach - scanning everything and dumping a mountain of alerts on engineering doesn’t work anymore. If you drown teams in tens of thousands of findings, one of two things happens: they stop shipping, or they ignore security. Both are catastrophic outcomes.

Varun’s view is that the future is precision and remediation. Fewer false alarms, more meaningful signals, and systems that help fix issues inside the workflow - not as an afterthought at the end.

In an AI-driven world, security can’t be a tax. It has to be a default.

Leadership, said plainly: what founders should hear early

I asked Varun what he’d tell a first-time founder who has an idea, term sheets, and momentum.

He didn’t start with product. He started with life.

First, make sure your family is genuinely on board. Startup building isn’t just a professional commitment - it’s a mental occupation. Even when you’re home, the company follows you. The people closest to you feel that, and you can’t talk your way around it. You have to align early, honestly, and explicitly.

Second, treat the first hires like the company’s DNA, because that’s what they become. Varun deliberately avoided building a “clone army” from his past companies, even though he could have hired faster. He wanted diversity of thinking and experience, because speed without depth creates fragile cultures.

Third, stop romanticizing stealth. Varun believes building in the open - talking to customers, iterating your message, refining the problem - is the only way to stay aligned with reality. And he returns to the line that frames this entire edition: if someone can hear your idea and beat you, the issue wasn’t the idea leaking. The issue was your execution.

The closing thought I’m taking with me

Varun didn’t describe success as a payout. He described it as building something enduring.

He’s arguing for a founder mindset that feels almost countercultural right now: greatness first, leverage second, outcomes last.

Build the company you’d be proud to run even if nobody ever bought it.

Because if you build it that way, you’ll be the one holding the cards when the options show up.

If there’s a single sentence that sums up this conversation, it’s this: the hard thing isn’t raising money or getting acquired - it’s building something great without taking shortcuts.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with David Hornik, founding partner at Lobby Capital and one of those rare venture investors who’s seen enough cycles to tell you what’s actually durable.

David has backed companies that became categories - Splunk, GitLab, Fastly, Ebates - but what made this conversation valuable wasn’t a list of wins. It was the operating system behind them: the belief that venture isn’t a transaction business, it’s a relationship business and that the best investors (and founders) win by caring deeply about the problem, not just the upside.

If you’re building, raising, or hiring in 2026, this one is a masterclass.

Let me walk you through what stood out.

Venture isn’t a deal business. It’s a people business.

David said something that should permanently reframe how founders think about fundraising:

A lot of people believe venture is about finding and doing deals.
They’re wrong.

In his view, venture is fundamentally a relationship game because companies are built by people, destroyed by people, and accelerated by people.

You can have a great product and still lose if the team is dysfunctional.
You can have a mediocre business and watch it become massive if the people are exceptional.

That’s also why “AI, tech, tech, tech” isn’t the full story. The tech matters, of course, but the human dynamics matter more. Especially when things get hard (which they always do).

And I loved the blunt truth embedded in his line: “There’s no great business on the planet that bad people haven’t ruined.”

It sounds obvious - until you’ve seen it happen.

The best VCs don’t just fund you. They sell you.

When I asked David what value a VC brings besides money, he didn’t reach for the usual “network” cliché.

He called it what it is: selling.

A great VC becomes your “cheerleader in chief” - not in a fluffy way, but in the most practical way:

When you’re trying to hire someone extraordinary - someone who has options - credibility matters. And a strong investor lends that credibility.

“David meets 1,000 companies a year and funds one. We were the one.”

That sentence sells.

It sells to candidates.
It sells to customers.
It sells to future investors.

But the piece that felt even more true - especially for founders - is what he said next: the best investors also become the place where you can tell the truth.

Because being a CEO is lonely.

You can’t always dump fear, doubt, or uncertainty onto your team - everyone reports to you.
So a good investor becomes part psychologist, part operator, part board partner - someone who can help you keep the business from eating you alive.

That’s not talked about enough. But it’s real.

A different kind of edge: logic + dyslexia

David has an unusual background: he was a lawyer before becoming a VC, and he’s dyslexic.

It’s a combination that sounds contradictory - until you hear him explain it.

Law trains you to think in rules, implications, risks, and tradeoffs.
Not just “what’s written,” but what happens if you don’t comply.

Dyslexia, on the other hand, forces a different skill: you rarely have perfect information, so you learn to extrapolate and make decisions with incomplete data.

And that’s basically venture.

You’re never investing with certainty. You’re investing with fragments.
So the ability to reason clearly and decide without full visibility becomes a competitive advantage.

It also connects to something founders forget: detail matters - but interpretation matters more. Great decision-making isn’t just about knowing the facts. It’s about understanding what the facts imply.

Fundraising leverage flips the moment there’s a term sheet

This was one of those “write it down” moments.

David told me a story about Travis Kalanick raising Uber’s Series A, and he shared a dirty secret most founders don’t internalize early enough:

Until a term sheet exists, the investor has leverage.
The moment a term sheet exists, the founder has leverage.

Because a VC spends a year looking at hundreds or thousands of deals to write one term sheet.
Once they’ve written it, they want it to happen.

That’s why the best founders slow down right when they feel pressure to speed up.
They create space. They compare. They negotiate from strength.

And the parallel to recruiting hit me immediately: the leverage shifts the moment intent becomes clear and closing becomes a two-way sale.

Seed is promise. Series A is proof.

David gave the cleanest framing I’ve heard in a while:

Seed rounds are mostly promise.
Series A requires proof.

At Seed, someone might fund you because you’re sharp, the story is compelling, and the market feels big.

At Series A, you need to demonstrate that the thing you believed is true:

● The problem is real

● The customer cares

● The solution actually solves it

● And ideally: people will pay for it

No shortcuts. No cheat codes.

This is the heart of PMF. It’s not storytelling. It’s validation.

And it’s why so many “high potential” startups stall: they confuse early excitement with durable demand.

Founder-market fit is overrated. Obsession with the problem isn’t.

This part matters because it cuts against the current founder narrative.

When I asked David whether he indexes on founder-market fit, he said it’s overblown.

Yes, sometimes the fit is obvious like a doctor fixing information flow in hospitals.

But he’s also backed founders who had no “obvious” right to win - young founders, outsiders, people who simply saw something broken and worked relentlessly to understand the customer and solve it.

What he does look for is more interesting:

He wants people who genuinely care about the problem - not people chasing what they think will make money.

And he wants founders who don’t pretend they have all the answers.

If you give confident answers to unanswerable questions, it’s a red flag.
The best founders have strong convictions where they should and real curiosity where they don’t yet know.

That’s a very different filter than “ex-Google” or “10 years in industry.”

The AI era is accelerating traction… and accelerating death

David’s take on AI companies was nuanced and honest.

Yes - teams are getting traction faster with fewer people.
Yes - products can be built quicker than ever.

But he’s worried about what comes next: defensibility.

In a world where everyone has similar tooling, the question isn’t “can you build it?”
It’s “can you defend it?”

Historically, the slope up is often the slope down.
Momentum attracts competitors. Speed invites clones. Differentiation gets thinner.

So AI will create a small number of massive winners - but it will also create an even larger graveyard of fast starters who couldn’t sustain a moat.

That’s the uncomfortable truth behind this era of “four engineers to $50m ARR.”

LPs don’t fund stories. They fund access.

David’s breakdown on fundraising for VC funds was also surprisingly applicable to founders.

LPs care about two things:

1. Proof you’ve backed breakout companies
   

2. Deal flow - will you even see the next breakout?

Because you can’t do great deals unless you see them.
And there are only a small number of transformative deals in any given year.

That mindset maps perfectly to founders and operators too:

You’re only as good as the next distribution edge.
The next customer insight.
The next hiring unlock.
The next relationship that changes the trajectory.

The deeper thread: giving wins

David ended by recommending Give and Take and The Biggest Bluff, and the theme was consistent with everything else he said:

People who give - who build relationships without keeping score - end up with the strongest networks, the best deal flow, and the highest-trust partnerships.

Not because it’s moral.
Because it compounds.

And that, to me, is the quiet meta-lesson of the whole episode:

In a world obsessed with speed, leverage, and tactics - relationships are still the highest ROI asset.

If you want one line to take from this edition, it’s this:

You can’t sell Series A on promise. You have to prove the truth.

Everything else - the fundraising dynamics, the investor-founder fit, the recruiting parallels, even AI’s acceleration - hangs off that.

And if you’re building right now, that’s the job.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode on Inside the Silicon Mind with Rob Bearden, co-founder and CEO of Sema4.ai. Rob is one of those rare enterprise operators who has lived through multiple cycles of building: co-founding Hortonworks, leading Docker, and then returning to Cloudera to take it through a major turnaround and ultimately a $5.3B sale.

What made the conversation unusually valuable wasn’t a single tactic. It was the way people, customer outcome discipline, and the AI platform shift all tied into one continuous PMF story. You can’t separate them. The quality of the team shapes the speed of PMF, the quality of customer outcomes shapes the durability of PMF, and the platform shift determines whether PMF survives the next wave.

Let me walk you through what stood out.

The people lens: why Rob starts with talent (every time)

Rob didn’t hesitate when I asked what mattered most across Hortonworks, Cloudera, and now Sema4.ai.

His answer was immediate: people.

Not “hiring is important” in the abstract. He meant something more specific and more operational:

Having the right people in the right roles for the phase the company is in is the difference between scaling and stalling.

The hidden PMF lesson here is that product-market fit isn’t only a product problem. It’s a phase-fit problem. The team that gets you to early traction is rarely the same team that gets you to repeatability, and the team that gets you to repeatability is rarely the same team that gets you through public-market rigor or a major exit.

Rob’s point was simple: you can’t “willpower” your way through phase changes. You need the right operators at each stage and leaders who can recognize when the company has outgrown the old operating system.

The customer outcome lens: “PMF is won by doing whatever it takes”

Rob kept coming back to one phrase that, honestly, should be written on every early-stage founder’s wall:

Be maniacally focused on customer outcome success.

He described a culture where the internal standard is not “ship features” or “hit deadlines.” It’s:

Did the customer get the outcome - predictably, repeatedly, at a level they trust?

And he took it further: if you need to lose money on a deal to ensure the customer outcome is successful, you do it - because the compounding effect of trust is the real engine of durable PMF.

The PMF lesson buried in that is uncomfortable but real:

PMF doesn’t come from persuasion.
PMF comes from outcomes that customers can’t unsee once they experience them.

When the outcome is undeniable, the buyer stops “evaluating” and starts “standardizing.”

SaaS to agents: why the build cycle just got compressed

Rob made a distinction that matters for every founder building today:

SaaS was a waterfall rhythm.
Agents are an iteration rhythm.

In the SaaS era, you could define a roadmap, build for 6-10 months, release, test, then roll to production. In the agent era, Rob described a different clock speed:

Week-long cycles, with customers experimenting in real environments, tuning for accuracy, and getting to meaningful production workflows inside 45-90 days.

The takeaway is not “move fast.” Everyone says that.

The takeaway is that the market is now structured so that:

If you can’t build and prove value inside compressed cycles, you don’t lose later - you get disintermediated early.

Which leads to his next point.

The “have vs have-not” divide: production value or clever tooling

Rob described a real split in the AI market:

● Some products are impressive, but still searching for a production-grade use case.

● Others are applied in a way that creates true value unlock inside an enterprise workflow.

And his definition of value wasn’t vague. It was outcome-based:

Value is when the customer can do work faster, more efficiently, and more accurately than they could with humans or constrained SaaS workflows - ideally with less human intervention.

This is one of the cleanest PMF tests I’ve heard for the agent era:

Is your AI doing something “interesting”?
Or is it delivering an outcome that the enterprise can trust enough to operationalize?

What Sema4.ai is really building: the agent platform the enterprise can standardize on

Sema4.ai’s mission, as Rob described it, is to be the platform for enterprises and ISVs to build, deploy, and manage purpose-built agents across mission-critical work.

The key is not “one agent.” It’s lifecycle management:

● building agents from natural-language descriptions of work and outcomes

● connecting to the right data and applications

● deploying with governance, security, auditability, and lineage

● scaling from a few initial use cases into many

He also shared something important about adoption dynamics:

Early adopters start with 1-3 use cases.
Once they see the value unlock, they realize they have 10 more and then the platform becomes the standard.

That’s the compounding motion that looks like PMF in the agent era:

Not “seat expansion.”
Use-case multiplication.

Why AI adoption is “inside-out” and why expansion is the hard part

Rob gave a great frame that explains why so many AI products feel like they’re “landing” but not scaling.

He described SaaS as outside-in:

● land is harder

● expansion becomes easier once the product is embedded

And AI as inside-out:

● land can be easier (a pilot, a use case, an early win)

● expansion is harder (because trust, governance, and controls become the bottleneck)

The PMF lesson is sharp:

In the agent era, winning a pilot is not PMF.
PMF is when the enterprise trusts you enough to expand across workflows.

The barrier every founder must take seriously: security, governance, and lineage

This was one of Rob’s most enterprise-native insights.

He argued that security isn’t separate from governance anymore because agents:

● access data through multiple steps

● shift context during execution

● create outcomes that may change who should be allowed to see what

So the enterprise question becomes:

1. What data did the agent access?

2. Who had permission?

3. What’s the audit trail of reasoning and execution paths?

4. Can we prove lineage end-to-end?

Rob’s answer was also practical: you don’t invent a new permissions universe. You operate within the customer’s existing permission frameworks which adds a “multi-dimensional engineering” challenge beyond the agent itself.

In other words:

The simplicity users feel is masking enormous complexity underneath.

And that’s why the winners won’t just be the teams with the best demos. They’ll be the teams with the strongest trust architecture.

The next role shift: outcome-based engineers

Rob made a prediction that connects directly to how PMF will be built:

We’ll still have software engineers, but we’ll also see a new class emerge - outcome-based, agent-layer engineers.

A hybrid of:

● engineering fundamentals

● workflow and process thinking

● outcome/KPI definition

● vertical context (finance, tax, supply chain, etc.)

Because in the agent world, the product isn’t a feature set.

The product is the outcome.

Closing thought

If I compress the entire episode into one sentence, it’s this:

PMF in the agent era will go to the teams who build for customer outcomes first, iterate at a weekly clock speed, and earn enterprise trust through governance, security, and lineage that makes expansion inevitable.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

This week’s PMF Playbook comes from my episode with Ross Lazerowitz, founder & CEO of Mirage Security. Ross has built across product and security at companies like Splunk and Observe, and he’s now focused on one of the most uncomfortable truths of the AI era:

As technical defenses harden, humans become the path of least resistance.

What made this conversation unusually valuable wasn’t just the cybersecurity angle. It was the way three themes stitched together into one continuous PMF story: platform shift (AI-driven deception), category formation (human security), and founder reality (fundraising + endurance). You can’t separate them. The platform shift changes the attack surface, the new attack surface creates the category, and the founder’s ability to survive the emotional load determines whether the company gets to PMF at all.

Let me walk you through what stood out.

The platform shift: the “authenticity era” is over and it changes everything

I opened with a question that sounds dramatic but is rapidly becoming operational:

If AI can fake your voice, write your messages, and learn how to manipulate your emotions… is human authenticity already over?

Ross’s answer was blunt: it’s been over for quite some time.

His mental model is that we’re heading toward a world where “real vs fake” becomes less important than “verified vs unverified.” He referenced the idea that parts of the internet already feel simulated - AI-generated content feeding AI-driven engagement loops.

The PMF lesson buried in that is simple and slightly unsettling: when a platform shift makes deception cheap, trust becomes the scarce resource. And scarcity creates markets.

Founders don’t get to ignore this. If your product assumes that identity, intent, or messages are implicitly trustworthy, you’re building on a foundation that’s eroding.

The category lens: security is moving from network/app/data… to “human”

Ross’s origin story here matters. Early in his career he worked in a bank’s Security Operations Center - alert triage at massive scale and he watched enterprises get compromised constantly.

But the key insight isn’t “security is bad.”

It’s that over the last 20 years, the industry has made real progress on the classic layers:

● network and perimeter controls improved

● MFA became normal

● patching, detection, and visibility got better

● “big firewall and pray” matured into more modern approaches

So what’s left?

People.

Ross framed it cleanly: if technology controls keep improving, attackers will increasingly choose the path of least resistance and that’s human behavior under pressure.

The PMF implication: Mirage isn’t competing in “more cybersecurity tooling.” It’s competing in a new layer of security that becomes inevitable as AI accelerates social engineering.

Social engineering has evolved and AI turns it into an assembly line

Most people still think phishing is “a bad email.”

Ross’s view is that we’ve moved into multi-vector attacks: phone calls, SMS, email threads, voicemails and deepfakes are the next multiplier.

He gave the modern version of the attacker story: not a hoodie in a dark room, but often young, highly effective operators using persuasion and process weaknesses to get access - especially through help desks, password resets, and identity workflows.

This is what matters for PMF: the attack pattern isn’t technically sophisticated - it’s operationally effective.

So the winning products won’t be the ones that add another dashboard. They’ll be the ones that harden the real workflows where humans actually break.

Mirage’s wedge: don’t “train people” - pressure-test the system

Mirage’s approach isn’t to lecture employees.

It’s to simulate realistic social engineering attacks at scale and measure where the organization fails before attackers do.

Ross described Mirage’s AI doing what a human attacker would do:

● calling into a help desk

● impersonating a believable internal persona

● applying pressure

● triggering resets and access changes

● testing whether controls hold up in the messy real world

A design-partner story hit hard: in one simulation, the help desk removed a geo-restriction “to the world” within ~30 seconds of a call, in a panic. If they hadn’t slowed it down, Ross believes the attack could have gone very far.

The PMF lesson is brutally practical: you don’t know your controls until you test them under stress. Security theater disappears the moment you run real scenarios.

This is also why the product is sticky: once a customer sees that gap, they can’t unsee it.

Deepfakes: detection is the wrong hill to die on

Ross has a contrarian (and I think correct) position: he’s bearish on deepfake detection as a reliable solution.

His reasoning is straightforward: there’s no technical reason detection can’t become a losing game. If generators improve, detectors chase. And “mostly accurate” isn’t acceptable when the downside is account takeover or fraud.

So the framing flips:

Don’t detect fakeness. Verify realness.

He pointed to provenance-style thinking - cryptographic signing and chain-of-custody for content - where trust comes from source and history, not whether something “looks fake.”

PMF insight: when a problem becomes unwinnable at the detection layer, categories shift toward verification, provenance, and process design. The product moat becomes workflow trust, not model classification accuracy.

Fundraising reality: seed is vibes, and the feedback will drive you insane

Ross was unusually candid about fundraising - not the tactics, the emotional cost.

At seed, there are no clean metrics. So judgment becomes subjective. And the hardest part is that feedback is often inconsistent and not fully trustworthy because investors don’t want to burn bridges.

The PMF lesson for founders: don’t over-iterate on noise. You need a tight internal compass, or you’ll contort the company around contradictory opinions.

Ross also called out a mistake a lot of first-time founders make:

Investor marketing ≠ customer marketing.

Customers don’t care about your grand vision deck. Investors don’t care about product nuance in the way customers do. Mixing the two wastes time and creates confusion.

Founder-market fit: trust = intent + expertise

Ross believes founder-market fit matters especially in security.

When you’re selling into CISOs at massive enterprises, they’re not just buying software. They’re buying belief that you understand the threat model, the workflows, and what breaks in production.

His trust equation was essentially:

● Do I believe your intent is to help?

● Do I believe you have the expertise to solve this?

If either is missing, the deal doesn’t happen.

This is why Mirage’s origin story matters: Ross isn’t trying to “learn security on the job.” He’s building directly inside a domain where credibility is the entry ticket.

Closing thought

If I compress the entire episode into one sentence, it’s this:

As AI makes deception cheap and scalable, PMF will belong to companies that stop trying to “educate humans” and instead redesign the workflows humans operate inside - shifting trust from perception to verification.

That’s what Mirage is building toward: human security as a first-class layer, tested under pressure, built for the world that’s arriving - not the one we wish we still lived in.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/

So when Zakir explains how Censys stayed true to its technical thesis while customers, investors, and industry narratives pulled them in every direction, you lean in.

Because this is the part of building companies most people skip: not the idea, not the funding, not even the product. The part where you’re forced to decide whether you’re building what’s requested, or what’s needed.

And those are rarely the same thing.

The Two-Lens Problem: Customer Pain vs. Company DNA

Zakir frames the entire challenge through two lenses: the “what” and the “how.”

The “what” is the customer problem - what security teams are actually trying to solve. That requires deep listening: customers, prospects, domain experts, people with a view into where the market is going.

But the “how” is the company’s DNA - what you uniquely bring to the table and refuse to compromise on.

At Censys, that DNA is uncompromising: data quality and internet visibility at extreme fidelity. Not as an engineering best practice, but as a strategic pillar. Zakir makes the point bluntly: you can’t solve customer problems without elite data, and you can’t build elite data without staying anchored to the real problems.

So Censys built the organization to enforce it. One side of the house benchmarks and improves accuracy, latency, and coverage relentlessly. The other side lives with customers, extracting the real pain behind what they’re asking for. And the magic is in the bridge between them: constant, intentional collaboration.

That’s how conviction becomes operational.

From University Research to a Company the Market Kept Pulling On:

Censys started as a research project at the University of Michigan, built around a deceptively ambitious question: how do you scan the entire internet at scale?

That became ZMap - an internet-scale network scanner inspired by Nmap, but designed to operate across everything, not just a range. Initially, it served academic curiosity: patching patterns by country, insecure infrastructure, trust relationships across certificate authorities.

But the moment the industry got hold of it, the demand shifted.

Security teams. Pen testers. Threat hunters. Defensive VM teams. Even market analysts measuring cloud adoption trends across AWS, Google, Oracle, IBM.

Everyone wanted visibility.

So the company did what “good startup advice” tells you to do: pick one use case and go deep. They focused hard on attack surface management - shadow IT, perimeter exposure, internet-facing infrastructure - helping defensive teams answer the urgent question: what’s exposed, what’s risky, and what do we fix first?

But here’s the tension: while the paid product narrowed, a free community product remained wide open - anyone could query any asset on the internet.

And over time, that open surface kept growing… even without heavy R&D investment.

That growth was the early signal of something most founders miss: the market was telling them the platform wanted to exist.

The Data Was the Product (Even When Everyone Told Them It Wasn’t):

Zakir shared a critical detail: early advice from the outside world was almost universally the same:

“Leave the data behind. The data isn’t valuable. Nail one use case. You can’t do two.”

And that’s the trap. Because the advice isn’t wrong - it’s just incomplete.

What Censys saw over time was something more structural: every time they built a feature for one persona, another persona needed it immediately. Data quality problems raised in one area would later surface in a completely different workflow. The same infrastructure relationships kept reappearing across use cases.

Eventually, the realization clicked:

Every security persona has a slice of the internet they care about - first-party assets, third parties, supply chain dependencies, critical infrastructure sectors, adversary infrastructure. Different missions, same map.

So Censys didn’t pivot into “being a data provider.” They did something smarter:

They built a platform that lets every persona make sense of their slice - using a shared foundation of best-in-class internet visibility.

That’s not a shift away from focus.

It’s a shift into architecture.

The Product Lesson: Don’t Build What Everyone Asks For:

Zakir’s view on the product is refreshingly honest: if you build exactly what customers ask for, you end up with a messy hodgepodge - features without a coherent story.

The job isn’t just listening.

The job is pushing past the surface request to find the kernel underneath:

● What problem are you actually trying to solve?

● Why does this matter right now?

● What’s the pain you’re failing to articulate?

And yes, there are table-stakes realities. Splunk integrations aren’t exciting. ServiceNow isn’t glamorous. But you build them because that’s how production works.

The differentiation isn’t in the plumbing.

The differentiation is in the core problem - visibility, attribution, accuracy, relationships, real-time context.

Or as Zakir’s philosophy implies: integrations keep you in the game; foundations win the game.

PMF, According to a Founder Who Doesn’t Romanticize It.

When I asked Zakir when he knew Censys had product-market fit, he gave the answer I trust the most:

It’s not a switch. It’s not an endpoint. It’s not “you have it or you don’t.”

It’s repeatability.

It’s when you can describe the problem and buyers feel it viscerally:

● “Yes, that’s us.”

● “Yes, we’ve got budget for that.”

● “Yes, we didn’t budget for it, but now we have to.”

And even then, it doesn’t end - because security changes constantly. Five years ago, VPN endpoints weren’t the breach epicenter. Now they’re a leading initial access vector for ransomware groups. The product must evolve with the threat surface.

PMF is a moving target.

The only way to keep it is to keep earning it.

AI: Not “AI vs AI,” But Humans Superpowered by AI

Zakir’s stance on AI is pragmatic. He doesn’t believe we’re in a world where AI is “magically hacking unknown vulnerabilities.” He believes we’re in a world where AI accelerates what humans already do - faster, more targeted, more scalable.

Better phishing. More tailored social engineering. Faster reconnaissance. Wider democratization of attacker sophistication.

On the defense side, the opportunity is equally clear:

● help junior analysts see patterns sooner

● connect disparate signals across tools

● automate the tedious work that drains security teams

● gradually build trust toward safer automation (including remediation)

But he’s candid about reliability. Hallucinations are real. Bad outputs happen. Trust isn’t binary - it’s earned through controlled use cases and gradual gating.

The first wave is augmentation.

The later wave might be automation.

First Principles: The Skill That Becomes More Valuable as Code Gets Easier

One of the most useful parts of the conversation was Zakir’s answer to a market narrative I hear constantly: “AI will eliminate junior engineers.”

He doesn’t buy it.

What AI replaces is the tedious work - the scaffolding, the templates, the glue code, the searching through unfamiliar codebases. But the hard problems remain:

● designing abstractions

● scaling systems

● testing boundaries

● building secure interfaces

● making tradeoffs with performance and reliability

● evolving systems without breaking them

That’s first principles thinking: not a buzzword, but the ability to understand what’s invariant beneath the tools.

And in a world where AI can generate code, the human value shifts upstream: design, judgment, architecture, and intent.

The Future of Security Is Less Dashboard, More Data Platform

Zakir described something I think every security founder should tattoo onto their roadmap:

Why is every security company reinventing Tableau?

Security organizations have dozens, sometimes hundreds of tools. Historically, everything dumps into a SIEM and humans stitch meaning across “15 tabs open.”

What’s changing now is bidirectionality. Customers don’t want Census to be a one-way data source. They want control over:

● how assets are discovered

● how ownership is defined

● how external signals inform internal prioritization

● how tools speak to each other through newer interfaces (like MCP-based APIs)

This is a profound platform shift: from black box product to extensible system where customers can interrogate decisions, view evidence, and integrate insight into their own ecosystems.

Less “here’s our dashboard.”

More “here’s the truth; use it your way.”

Closing Thoughts

Zakir’s story is a blueprint for founders building deep tech in noisy markets.

He didn’t win by chasing every request.

He won by holding a technical line: data quality, fidelity, real-time visibility and letting use cases accumulate on top of a shared foundation until the platform became inevitable.

It’s a reminder that the best companies don’t just listen.

They translate.

They interpret.

They decide.

And then they build with enough conviction that when the market finally catches up, it looks obvious in hindsight.

Until next time,

Firas Sozan
Your Cloud, Data & AI Search & Venture Partner

Find me on Linkedin: https://www.linkedin.com/in/firassozan/
Personal website: https://firassozan.com/
Company website: https://www.harrisonclarke.com/
Venture capital fund: https://harrisonclarkeventures.com/
‘Inside the Silicon Mind’ podcast: https://insidethesiliconmind.com/