What’s the deal with Zuckerberg and AI?

A brief timeline of Meta and AI in 2025

May 27, 2025

Exclusive: Meta shuffles AI team to compete with OpenAI and Google

Axios reports that Meta restructures their AI division into an AI Products team and an AGI Foundations team. The former focuses on AI in Meta’s services and the latter on developing LLMs and better capabilities in reasoning, multimedia, and voice.

Yann LeCun remains at Meta’s AI research unit FAIR (Fundamental AI Research), which is focused on developing future AI … eh… “algorithms”, for lack of a better word.

June 13, 2025

Meta poaches 28-year-old Scale AI CEO after taking multibillion dollar stake in startup

Reuters reports that Meta invested $14.3B in Scale AI and recruited its CEO, Alexandr Wang, as Chief AI Officer. Wang now leads the new superintelligence unit as Meta aims to revamp its AI strategy.

August 16, 2025

Meta plans fourth restructuring of AI efforts in six months, The Information reports

Meta divides its AI unit, Superintelligence Labs, into four groups: a new TBD Lab, a products team including the Meta AI assistant, an infrastructure team, and FAIR.

August 21, 2025

Meta freezes AI hiring, WSJ reports

After hiring about 50 new researchers, Meta pumps the brakes on acquiring new talent.

August 26, 2025

Researchers are already leaving Meta's new superintelligence lab

WIRED has confirmed that three recent hires have now resigned.

I said hey…

That’s a lot of things happening in a short timeframe. Full speed ahead in late spring, followed by a hiring freeze and multiple restructures in late summer. So what’s the plan? Is there a plan? Let’s explore three options:

Move fast and break things

This is the OG Facebook style: throw things at the wall and hope none of it is the kind of shit that hits a fan. Basically, Meta is using its money reserves to try and luck its way into a successful AI strategy. There is no plan other than trying as much as possible in as short a timeframe as possible.

Historically, Meta has failed at building anything successful beyond ‘The Blue App’, but Zuckerberg has a long history of copying or buying other services that have strong growth potential. There’s no doubt that he’s noticed that all the money in tech is flowing towards AI, and no doubt that he wants some of that. Cue going on a spending binge.

Clever ploy

Meta has one thing that no other AI company has: profits. OpenAI, Anthropic et al are burning money like there's no tomorrow, and are dependent on ever-wider circles of investors handing them money. What they do have, and what Meta lacks, is a stable of well-known AI researchers. So the obvious play is to leverage what you have as a weapon to weaken your opponents’ strengths.

Yes, Meta seems to have largely failed to recruit people, even with the obscene bonuses that have been rumored.

But it’s probably safe to assume that Zuckerberg hoped that the people Meta targeted got more than a bonus week off in order to stay. OpenAI etc probably had to raise a few salaries (whatever Dario Amodei says), and hand out a bunch of stocks in order to keep people from jumping ship. That means they have to spend even more money than before, money they don’t have, so they’ll have to go begging for handouts sooner than expected.

In this scenario, Zuckerberg is treating unprofitable companies like opponents in a game of Civilization. Trying to starve them out so that he can pick up the remains.

Panic Mode

In late October 2021 Facebook went all in on “The Metaverse”. Anyone with a clue saw it for what it was, a diversion from all the horrible stuff that then Facebook had caused. A desperate attempt to seem like a growth company rather than an established behemoth.

That didn’t age well. After stumbling around and stepping on every rake within walking distance, from VR goggles to turning Instagram posts into NFTs, even the eternally clued out Zuckerberg seems to have realised that he needs to look as if he’s part of the AI boom.

In an age of ridiculous valuations for anything backed by an LLM, why not shout from the rooftops that you’re “DOING AI!” and promise to build a datacenter the size of Manhattan?

The difference between this and the “Move fast” scenario is that here Meta is blindly reacting, while in the former they were blindly acting. If that makes sense.

What’s going on!?

Clearly some people at Meta are well aware that LLMs are a dead end on the path to actual intelligence. But that’s not how you make money in Silicon Valley. Bold proclamations of future inventions is the way to go.

So having a large part of your company making a lot of noise, while the ones with a shot at creating something revolutionary can tinker in the background is not a bad way to go about things. Not that I think this is the plan, looking at the Metaverse fiasco tells me that the chances of a multipronged approach with intentions behind it is well beyond Meta's capabilities. But the end result might be some peace and quiet for YannLeCun to pivot into whatever the next thing after Transformer models might be, while Mark can appear as strong and intimidating as Musk found him during the whole cage fight incident.

Tactics without strategy…

It’s been two and half millennia since Tsun Zu wrote down some pretty basic tips regarding the art of war, one of the best known ones is: “Tactics without strategy is the noise before defeat“.

And yet here we are… with Meta repeatedly acting without any long term thinking, constantly flushing money down the drain. We’ve seen the effects on their attempt at a stable coin, their metaverse pivot, and Zuckerberg’s perennial wardrobe changes. All best summed up as #fails.

I predict that in two or three years' time we’re going to see the next hype build around some other vaguely defined thing that demos well. Two or three years after that, and Meta will probably pivot as hard as they can towards it. See you in 2030.