Sam Altman's 10 Rules for the AI Era

The VC Corner | Ruben Dominguez | May 2026

---

1. Key Themes

Theme 1: AI Infrastructure Is Being Built at Civilizational Scale — And the Demand Is Effectively Uncapped

OpenAI's 20-year infrastructure commitments signal a level of conviction that most operators and investors haven't internalized. The underlying logic is a demand curve that runs opposite to most industries.

"In some sense, I think demand for intelligence at a low enough price is effectively uncapped."

Every time the price of AI drops, new use cases emerge that didn't exist at the prior price point. The infrastructure is already committed — the question is what gets built on top of it.

---

Theme 2: The Founding Filter Has Fundamentally Shifted — Domain Experts Are Now Fundable

For two decades, no technical co-founder meant no deal. That filter has changed. Altman now explicitly wants to back founders with deep user understanding who can't write a line of code, because the model itself is the "guitar player."

"There was a time when we used to make fun of the idea guy... it would be like saying, I have a great idea for a song. I just need that guy with the guitar to make it for me."

Operators, clinicians, lawyers, and domain experts who were previously unfundable now have a short-window arbitrage opportunity — most of them don't know it yet.

---

Theme 3: AI Will Compress a Decade of Scientific Progress Into a Single Year

The article flags this as the most consequential claim Altman made in the full two-hour conversation — and the one that received the least attention in the room. Three specific verticals are named: biology foundation models, complex disease drug discovery, and materials science.

"If we can start doing a decade of science of what it would have taken us in the old world in a year, the compounding effect there and what we'll be able to do and discover will just be extremely great."

The article notes that most investor attention is on apps and agents, while "the civilizational leverage is in science" — making these sectors structurally underpriced relative to their potential impact.

---

Theme 4: The "Hoping AI Gets Smarter" Test Is the Universal Product Filter

A single diagnostic question can resolve every "is this a real company" debate faster than any due diligence process.

"You as a business want to be on the side of hoping that AI gets smarter."

If a model capability improvement removes the reason your product exists, you were renting a gap — not building a company. If your product gets materially better every time a new model drops, you pass. If you're quietly hoping capability stalls, that's the clearest possible signal you're on the wrong side of the curve.

---

2. Contrarian Perspectives

Perspective 1: Keeping AI in the Lab Was Never About Safety — It Was About Power Concentration

The consensus view pre-ChatGPT was to keep AI inside research institutions, framed as a safety argument. Altman rejected this on a structural ground that inverts the conventional framing entirely.

"It is extremely important that we avoid that kind of power concentration and that we build this for the world."

The argument: locking transformative technology inside a small research cohort creates the concentrated power structure that safety rhetoric claimed to prevent. Iterative public deployment was the contrarian call — and it was a closer decision internally than is commonly understood.

---

Perspective 2: OpenAI Deliberately Chose Low Margins — And Considers That a Feature, Not a Bug

The dominant assumption is that AI incumbents will extract high margins as moats solidify. Altman explicitly rejects this model, benchmarking OpenAI against Stripe rather than Google.

"I'd be happy for us to be a forever low margin as long as we can be huge and growing fast business, and I would like us to supply kind of an intelligence meter."

The strategic logic: switching costs in AI tools are lower than most assume, so lock-in through capability is the only defensible long-term position. As models improve across the industry, margins compress for everyone — making volume and customer alignment the durable position. Founders building on OpenAI's infrastructure should take the stated intent seriously: an infrastructure provider that competes with its customers destroys the trust the infrastructure depends on.

---

Perspective 3: Materials Science Is "Massively Underrated" and Most Portfolios Are Ignoring It

While the investment community clusters around AI apps, agents, and infrastructure software, Altman specifically flags materials science as an overlooked compounding bet.

"Materials science. Altman calls this massively underrated. New materials unlock physical limits in energy, computation, and manufacturing simultaneously."

The article frames it explicitly: "Most portfolios are making only one of the two bets on the table" — competing for the same agent infrastructure deals while leaving the science acceleration layer largely uncontested.

---

3. Companies Identified

---

4. People Identified

---

5. Operating Insights

Insight 1: The CEO Must Go First — Personally — With No Leaderboard

The Shopify playbook works specifically because Tobi Lutke built AI automations himself before asking his team to. The behavior propagates when the leader goes first; it stalls when the leader delegates.

"It was not like a token leaderboard. It was not some other kind of gamified hackable thing... the CEO of the company said, we are now going to put AI into everything we do."

The tactic: Build one workflow automation yourself this week using Claude or ChatGPT. Your team is watching to see if you mean it before they believe it.

---

Insight 2: Shared Conviction — Not Process or Incentives — Is What Makes Elite Technical Teams Function

The single management skill identified as the true driver of OpenAI's breakthroughs was Altman's ability to hold together a team of people who each believed they were the most capable person in the room.

"You figured out how to get a lot of people who all thought they were the only capable or most capable person, and everything had to go their way, to work together long enough to figure out the breakthroughs. That was the magic of OpenAI."

Three conditions made it work: concentrated resources on one direction, shared conviction in scaling laws before the market confirmed them, and a mission large enough that the pain of collaboration was worth it. Lose the shared belief, and the pain ends the company.

---

Insight 3: Audit Your Own Drudgery Before You Build a Product Around It

The next major wave of AI adoption won't be a new industry — it will be the sudden recognition of how much time people spend on tasks that shouldn't exist in their workday at all.

"The degree to which most people will realize they can sit back and watch an AI do most of their drudgery is going to surprise people."

Altman reports this firsthand: automating his own drudgery didn't just save hours — it changed the subjective quality of his work. The market hasn't priced this correctly because the category stays invisible until it disappears.

---

6. Overlooked Insights

Insight 1: OpenAI Finished Training GPT-4 Eight Months Before Launch — and Operated With Zero External Feedback

This detail is buried inside the "collective psychosis" anecdote but carries significant strategic weight. The team used GPT-4 daily for eight months with no market validation, no external signal, and no confirmation that their conviction was grounded in reality.

"We'd walk the halls sometimes. Are we engaging in collective psychosis?... And there was no feedback to keep us in check or sane from the outside world."

For founders navigating a genuine technological discontinuity, this reframes what "conviction" actually requires — not confidence, but the ability to function without external confirmation for extended periods. Most founders and investors are not structurally prepared for that experience.

---

Insight 2: OpenAI Is Piloting a Program to Embed Full-Time Employees Directly With CEOs

Mentioned once and not elaborated on, this is potentially a significant go-to-market and enterprise adoption signal.

"OpenAI is now testing a program where they embed a full-time employee directly with a CEO to automate that CEO's own workflows first. The fractal effect follows naturally."

If OpenAI scales this model, it represents a direct sales and adoption motion targeting the C-suite — bypassing IT procurement entirely and seeding adoption from the top of the org chart down. This could become a replicable playbook for any AI company selling into enterprises where executive behavior is the rate-limiting variable.