Few days ago, WSJ reported that Thinking Machines Lab (TML) Co-Founder Andrew Tulloch left the startup that has just raised a $2bn Seed on $10bn valuation, to re-join Meta’s AI lab for a package of up to $1.5bn (!)

The story of top AI startup founders getting poached by big tech has become a pattern. Scale AI, Windsurf, Adept, Inflection AI… the list goes on - and the mechanism behind it seems the new way of M&A in the era of AI.

But what does that mean for employees and investors who get left behind in the original startup company?

How can investors prevent a departing founder to tank their investment?

Today’s piece is a reflection of what’s going on and what investors can do about it.

Let’s dive in

In the AI arms race, every startup with a world class research team becomes an acquisition target. Great exit channel and downside protection for investors.

True - but at the same time a big risk as shown by above examples where key researchers got poached and the oftentimes empty shells with shareholders and not so hot employees got left behind. Sorry!

It’s a double-edged sword that requires deeper reflection.

First of all, every acquirer faces the question:

👉 Do you buy the stock, or just the assets?

Buying stock means inheriting all obligations across equity, debt, contracts, liabilities.

Buying assets, on the other hand, lets you cherry-pick what you want: the people, the models, the brand.

So from the buyer side, the obvious choice in a war for AI talent is poaching key researchers instead of buying their companies.

In startups, especially deep-tech and AI, the moat often isn’t data or models. It’s the founder(s).

When that one critical founder leaves, so does the DNA of the company. The best “protection” against loss is simply that the founder never quits. But if they take a better offer - perhaps a lucrative acquihire, or a leadership role at a hyperscaler - all security is gone overnight.

So, how can we mitigate this founder risk?

How can we prevent founders to leave for the next best offer?

When evaluating the hottest AI labs, the due diligence shouldn’t just focus on data access, model performance, or commercials - it should focus first and foremost on founder durability.

Key questions investors should ask:

Said differently, do they view this as a multi-period game - one where reputation compounds and people meet multiple times in life - or a single-turn game, where they’re willing to burn bridges and pay with their reputation for a life-changing quick win?

Investors should spend even more time with the founding teams in cases where majority of the conviction and security rests on people. Sounds intuitive but if I look at some of the hottest AI rounds that I’ve seen first-hand, I’m honestly surprised how few investors took time to meet the founders in-person, went out for dinner & drinks, and learned about their personal life.

FOMO eats brains, sometimes.

What eventually matters most is the following: Founders who play single-turn games may look like bold opportunists but often leave a trail of scorched earth behind them. Multi-period thinkers - those who know they’ll build again - treat investors, employees, and acquirers as long-term partners.

The best founders - and the best investors - treat venture as a repeated game. Every decision today affects your ability to play tomorrow.

In a single-turn world, founders take the money and run.

In a multi-turn world, they protect their reputation as fiercely as their cap table.

Our goal as investors should be to back multi-turn founders who take serious responsibility, not short-sighted single-turn opportunists who don’t give a f*ck about the people who invited them to the dance in the first place.

Share this article with others who might benefit.

Good question. We all know the mechanics of tying founder incentives to the success of their business: Vesting schedules, lock-ups, tag alongs, liq prefs & more.

They work well, as long as founders play the multi-period long-game.

They fail as soon as founders opt for the single-turn option.

I have yet to see a suitable mechanism that would’ve prevented the Scale AI, Windsurf, or TML poaching cases. As soon as a founder accepts to lose everything in favour of a life-changing over, legals won’t protect you anymore.

The recent wave of AI M&A reminds us:

In the end, the true defensibility of any frontier AI startup isn’t in its model weights or valuation — it’s in the psychology of its founders.

Stay driven,
Andre