Mike Vernal on what Facebook taught him about patience — and why big tech is too risk-averse to fix email

Summary

Mike Vernal spent 15 years operating — mostly at Facebook, briefly at Microsoft before that — before becoming an accidental investor at Sequoia a decade ago. He now invests at Conviction alongside Sarah Guo, with a portfolio that includes Rippling, Notion, Clay, and Verata.

Patience as competitive advantage

The sharpest lesson Vernal carries from Facebook is patience, specifically the kind Zuckerberg showed — bias toward action in the short term, but willingness to hold conviction until new data proves the thesis wrong. He cites Reality Labs, where Facebook has spent roughly $80 billion, as the clearest expression of that posture.

The pattern shows up across his best investments. Figma, Notion, and Clay each took four to five years before breaking out. Airtable was founded in 2010 and didn't break out until 2017 or 2018. Vernal sees financial-only investors as systematically underweighting this — he describes a board member pressing a founder on why they weren't yet at $5M ARR, with the competitor already there. His read is that investors who haven't built something struggle to sit with the messiness of early-stage iteration.

He draws a distinction between stubborn vision and flexible tactics, using Notion as the example. Ivan Zhao's core idea — a canvas with composable blocks — hasn't changed in 12 or 13 years. The 2013 product demo, which Vernal says is findable online, already had the blocks, the canvas, the Stripe integration. What changed was everything around execution: the product was too complex, so the team stripped it back to a great note-taking app and rebuilt from there until something caught.

He's also direct about the limits of the patience framing. Founders invoking Figma as cover for slow progress don't earn that comparison automatically. Figma spent years in relative obscurity building something technically hard in a market where multiplayer design collaboration wasn't yet a trend. That window, Vernal argues, has to be earned.

Email and big tech risk aversion

Vernal has had a Gmail account for 21 years. His inbox should, in theory, be the ideal training set for a model that writes all his replies. Instead, he gets two-word completions. His conclusion is that Google is simply too risk-averse to build anything genuinely useful here — the downside scenario of an AI sending something offensive on a user's behalf is enough to kill the product internally.

The same logic applies to Siri, which he calls still pretty bad. The Meta AI assistant in the Ray-Ban glasses is an exception he considers actually good, but it requires wearing the glasses.

What he wants isn't summarization — he flags receiving a very long email as a negative signal in itself. What he wants is a model that reads the email and writes exactly what he would have written, especially on mobile, where composing is painful. He considers this a tractable problem, not a hard one, and is surprised no startup has shipped it credibly yet.

His broader argument is that horizontal AI experiences — the category email falls into — are now addressable by startups precisely because the incumbents won't take the necessary product risks.

Enterprise AI adoption

On which AI products are actually working in enterprise, Vernal names Cursor, Harvey, Decagon, and Sierra as clear breakouts. His framework for why some AI companies stick and others churn is straightforward: winners build down from a deep understanding of the customer; losers hypothesize what customers might need without having lived the problem. He expects that pattern to be the defining characteristic of the cycle.

On whether AI is sustaining or disruptive to existing software companies, he comes down clearly on sustaining in almost all cases. The one possible exception is Google. A competent incumbent that can still ship product, in his view, will adopt AI rather than be disrupted by it — the analogy he uses is that Salesforce for mobile was just Salesforce.

Consumer and M&A

Vernal ties the dearth of interesting consumer companies over the last decade directly to the regulatory crackdown on big tech acquisitions, which he dates to roughly 2016 and 2017. Consumer investing depends on a functioning exit market — very few things work, but when they do they work big, and the companies that don't work often have talented founders who can contribute inside a larger organization. Remove those acquisition paths and investors stop funding the category. His view is that consumer doesn't fully come back without either a major platform shift or a reopening of the M&A environment.

On consumer agents specifically, his framework is that the risk of direct competition with foundation model providers scales with how general-purpose the product is. Building for billions of people puts a startup in direct conflict with OpenAI. Building for hundreds of millions around a specific customer need is safer, because OpenAI can only put its best teams on the top 10 problems it has — anything outside that list is an opening for a startup.

Conviction

On why he joined Conviction rather than raising his own fund, Vernal applies startup logic to venture itself. The firms he respects most built from a focused position — Paradigm with crypto, Ribbit with fintech — and expanded from there. Trying to launch as a multi-stage generalist from day one is the equivalent of a startup announcing it's building the next Google. Conviction's thesis, focused on AI and intelligent software, follows the same playbook. He's been friends with Sarah Guo for nearly a decade — and recently discovered she'd sent him a LinkedIn message in December 2016 that he missed until two weeks ago.