Oren Zeev on why to stay small and solo, and the nature of AI investing today.

Samir calls the current market “disorienting”: some companies are growing faster than ever, while supply-demand imbalances are driving extraordinary private valuations. Oren says the disorientation is not necessarily bad because a completely clear market would leave fewer opportunities. He sees a sharply bifurcated market in which a small group attracts extreme attention and pricing, while the rest does not look like a bubble. He also sees a major private-public disconnect. Public-company multiples generally appear rational to him, whereas private companies are sometimes valued far above comparable public businesses; in those comparisons, he believes the public market is usually the better reference.

Samir frames the problem through the idea that the cost of missing a winner can be 50x, 100x, or 200x the cost of investing and losing 1x. Oren agrees that this drives FOMO, but adds a structural reason: a fund that has raised billions cannot generate strong returns through ordinary exits and therefore must own the very largest companies. The combination of large funds, social proof, and fear of omission creates a herd mentality around fast-growing companies. Oren expects both healthy and unhealthy growth stories to receive funding during frothy periods, with perhaps only one, two, or three eventual winners paying for the rest.

Samir describes two investor camps: those already invested in hyper-growth companies and effectively playing with house money, and those who missed them and now face pressure from their LPs to buy in at high valuations. Oren says both groups are reinforced into continuing the behavior, especially after the last three or four years made momentum investing look brilliant. He remains comfortable with omission mistakes when a deal falls outside his early-stage playbook, just as he accepts missing healthcare opportunities because he does not invest in healthcare. He also believes many less fashionable companies offer attractive opportunities because they do not fit the “could be a hundred-billion-dollar company” mold.

Oren characterizes AI’s first investment wave as infrastructure, models, chips, data, and data centers. He believes the application layer is only beginning and has not yet been overhyped, partly because application companies have experienced multiple contraction rather than expansion. Samir cautions that being right about transformative technology does not guarantee being right about where value will accrue. Oren’s test is whether a company is an AI beneficiary or victim, with no meaningful neutral category. He rejects the idea that OpenAI or Anthropic will simply replace every application, or that customers will “vibe code” all business software themselves.

Oren argues that application companies can build defensibility through domain expertise, proprietary data, customer trust, partnerships, and acquisition channels, even if the underlying technology is easy to reproduce. The code itself may be less of a moat because AI makes building faster, but regulated and high-stakes markets still require reliability: accounting, finance, and legal products cannot be correct merely 99% of the time. For existing SaaS companies, the outcome depends on whether leadership is willing to discard old assumptions, cannibalize the existing business, and adapt quickly. Oren trusts founders more than professional CEOs to make that transition because founders are less likely to preserve “sacred cows.”

Samir asks how investors should distinguish durable AI businesses from companies showing hypergrowth with questionable economics, citing reports that Harvey may spend $1.50 in inference costs to earn a dollar. Oren says current gross margins provide useful information: a company with high margins today is safer than one promising that costs will eventually fall, because competitors’ costs will fall too. A low-margin company can improve only if it develops a real competitive advantage. He is also cautious about AI companies expanding into services or bodies of work, since services have traditionally carried low margins and multiples. The key question is whether the product is genuinely sticky and hard to replace.

Samir worries that companies growing 2x, 3x, or 4x annually may be overlooked when other businesses are “speed-running” from $1 million to $10 million to $100 million and beyond. Oren rejects the idea that such companies are suddenly unfundable. If a company grows from $20 million to $60 million and then $150 million with healthy economics, he remains interested; five years of doubling produces roughly 32x growth. He says capital is still widely available for a $20 million or $30 million market leader growing 3x with real fundamentals, even if it cannot command 50 times revenue. Profitability alongside growth can also help.

Oren says he has an “allergic reaction” to doing what everyone else does. He believes he is more differentiated and more effective with his own early-stage playbook than by trying to play a large platform’s game, and expects better results for LPs from funds around $200 million rather than $1 billion. His larger funds raised around 2021 were more difficult vintages because more dollars went into highly valued companies and because a larger fund pushed him toward later-stage exposure. He does not claim large platforms are irrational: a $10 billion fund needs exposure to companies worth $100 billion, $200 billion, or even trillions to move the fund’s returns.

When Samir asks how Oren decides to follow on when other investors hesitate, Oren says later-stage conviction must rest on both the founder and the business numbers. He uses Navan, formerly TripActions, as an example: during the 2021 COVID period, travel-related metrics deteriorated, but the reason was visible and temporary. Oren and the company’s major investors believed the leadership was world-class and that business travel would return; they did not believe people would permanently stop traveling for work. He now sees Navan as an AI-enabled travel agency, with AI accelerating the product and creating no identifiable new risk. He says public markets still undervalue its story.

Oren compares AI with the internet: both are technologies that affect nearly every business and create opportunities beyond any single vertical. He sees differences from the dot-com era: today’s excitement is concentrated in genuinely fast-growing companies, and the private market is driving much of the pricing rather than the public market. He does not expect a repeat of the 2000 crash. His main anxiety is not business-related but societal: AI could affect democracies, antisemitism, and the job market, especially for people entering careers. He agrees with the risks raised by calls to slow AI, but doubts “pacing” is a practical remedy while countries compete; he hopes beneficial AI can counter harmful AI.

Looking back over three decades, Oren identifies chips and hardware as a missed opportunity from three or four years earlier, while calling several large proptech bets made before interest rates rose a mistake. He struggles to extract a universal lesson because hindsight makes decisions look clearer, but says investors should remain cautious without becoming cynical. Venture returns are concentrated in winners, so losing money on some investments is acceptable if the investor continues to believe in the possibility of exceptional outcomes. He and Samir agree that uncertainty is uncomfortable but productive: it creates risk and opportunity for both founders and investors.