Meta Paid Top Dollar for Jiahui Yu — He Still Left
Y•20h ago•editedJiahui Yu just announced that he’s leaving Meta to start a company, and the timing is pretty interesting.
He was one of the high-profile AI researchers Meta recruited from OpenAI and became a key figure around TBD Lab’s multimodal work. Public reporting also suggested Meta was offering extremely aggressive packages to recruit this tier of AI talent.
So seeing someone leave this quickly makes me wonder whether the bigger issue at Meta AI is no longer compensation, but the organization itself.
A few things I’m curious about:
-
If compensation is already extraordinary, what actually drives someone like this to leave? More research freedom? Faster execution? Ownership? Or simply much larger founder upside?
-
How stable is TBD’s direction internally? Meta’s AI org has gone through repeated restructurings and priority changes. For ambitious research teams, constantly shifting scope can matter more than compensation.
-
What does this environment look like for regular ICs? Top researchers can leave and start companies. Everyone else still has to deal with changing priorities, org reshuffles, scope competition, and projects that may suddenly lose sponsorship.
-
What exactly is the startup bet? If you already have one of the best-paid AI jobs in the industry and still choose to leave, you must believe the ownership/upside or ability to move faster is worth giving up a lot.
My current take is that Meta has clearly proven it can recruit elite AI talent with money.
The harder question may be whether it can create an environment where those people want to stay for five or ten years.
Money solves recruiting. It doesn’t automatically solve research freedom, organizational stability, ownership, or retention.
Anyone working around Meta AI / TBD have a different read on this?
Discussion
2 commentsSign in to join the discussion
Ask a follow-up question or share practical context.
The retention question is more interesting to me than the comp number. Meta can probably match almost any employee compensation package. What it can’t easily match is autonomy: choosing the research direction, hiring your own team, controlling the roadmap, and owning a meaningful percentage of the outcome. If more top researchers leave shortly after getting huge packages, I’d start looking much harder at org structure and research freedom than salary.
I wonder if this is less about Meta “failing to retain” him and more about how absurd the founder upside looks right now for elite AI researchers. Once you already have reputation, access to capital, and a team investors will fund almost immediately, even a massive Meta package may start looking like the *low-risk / low-upside* option. The interesting question is whether we’re entering a market where big tech simply can’t economically retain the very top 0.01% of AI talent long term.