AI Leaves The Lab, Lawyers Move In, Meta Bleeds Out

The Lab Door Is Open

Yann LeCun and collaborator Vert spent the day talking about getting AI out of the lab and into the physical world, which is the polite academic framing of a question MIT is asking more bluntly. Can these systems actually comprehend reality, or are they just very confident tourists in it? MIT Technology Review's framing was generous. The honest answer is that nobody knows, and the people building the robots have stopped waiting for one.

This is the structural tension of the week. The research side is still litigating whether large models grasp causality, object permanence, and the basic physics of a room. The industry side has already priced in the assumption that they will, eventually, well enough. Capital does not wait for epistemology.

Cursor's Exit, Workday's Reprieve

Cursor crossed a $3 billion run rate, an absurd number for a code editor wrapper two years ago, a defensible one now, and apparently a stepping stone to being acquired by SpaceX. Whatever your model of Elon Musk's empire, adding an AI coding company to a rocket and satellite business is a sentence that would have read as parody in 2022. It reads as Tuesday now.

Workday, meanwhile, posted strong results and the stock rose, which the market interpreted as proof that the AI apocalypse for enterprise SaaS has been delayed another quarter. This is the pattern. Every earnings cycle, an incumbent demonstrates it still has customers, and analysts conclude the disruption thesis is overblown. Every quarter after that, the disruption thesis returns. The companies that survive will be the ones that used the reprieve to rebuild, not the ones who took it as vindication.

Meta's Two Bills Come Due

Meta is having a worse week than its stock price suggests. Wired catalogued staff departures while Google reorganizes around AI with more conviction, a contrast that frames Meta as the giant losing its nerve. Separately, the company settled a social media addiction suit brought by a school district. One down, roughly 1,200 to go. The settlement amount was not the story. The story is that the addiction-litigation theory of social media harm has graduated from activist talking point to enforceable claim, and the AI-generated content flood Meta is currently encouraging on its platforms will feed the next wave of those suits.

Google doubling down looks decisive until you remember that decisive is also what Meta looked like during the metaverse pivot. Conviction is not a strategy. It is a mood.

The Copyright Front Widens

The Peanuts estate sued both AI companies and federal agencies over unauthorized use of musical compositions, which is a notable escalation because it names the government as a defendant alongside the labs. Most rights-holder suits so far have targeted private actors. Pulling federal agencies into the same complaint signals that catalog owners have decided the public sector's use of training data is not a free pass either.

On the other side of the same problem, Spotify and Universal announced the first major streaming deal to monetize AI-generated covers. Subscribers can remix licensed songs with AI, the rights holders get paid, and the platform takes its cut. This is the shape of the settlement most of the industry has been groping toward. License the training, license the output, route the money. The artists whose voices and styles end up in the dataset will, as always, be the last to see any of it.

The Peanuts suit and the Spotify deal are the same story told from opposite ends. Rights holders with leverage cut deals. Rights holders without it sue. The legal infrastructure for AI content is being built in real time, mostly by lawyers, mostly to the benefit of whoever already owned the catalog.

The Slot In One Line

AI is moving out of the lab faster than anyone has figured out what to do when it gets there, and the bills, legal, human, organizational, are starting to arrive in the same envelope.