Meta's FAIR Cull: 318 Menlo Park Names and a 30-Day Sourcing Window

Meta filed a WARN in California on Oct 22, 2026 listing 318 layoffs at its Menlo Park headquarters, the Bay Area slice of a 600-person Superintelligence Labs cut that hits FAIR, product AI, and infra. Terminations are dated Nov 21 in the SFGate reporting and Dec 22 in the EDD filing itself, and the roster spans two software engineering VPs, one AI research VP, and named ICs across search, voice, personalization, assistants, and robotics. If you recruit for OpenAI, DeepMind, Anthropic, or a well-funded lab, this is a 30-day window on a pool that will otherwise be absorbed into Meta's own TBD Lab or picked off by the biggest checks.

What the Oct 22 WARN actually says

Meta's WARN filing names 318 Menlo Park employees, spread across five office buildings with the majority at 1 Meta Way, with terminations dated Nov 21 per SFGate and an EDD-effective date of Dec 22. That gap matters: the Nov 21 date is when badges likely stop working and the pool becomes reachable on personal email; the Dec 22 date sets when payroll formally ends and severance clocks visibly begin.

Per the WARN and the memo from chief AI officer Alexandr Wang that same day, the composition breaks down like this:

• 318 layoffs in San Mateo County, concentrated at 1 Meta Way
• Cuts spread across software engineers, data scientists, privacy program managers, and AI researchers
• Teams named: search, voice, personalization, assistants, robotics
• Orgs affected: Superintelligence Labs, FAIR AI Research, product infrastructure
• 3 VPs on the list: two in software engineering, one in AI research
• TBD Lab, the small elite team building Meta's next-gen foundation models, was not affected

The Menlo Park 318 is 53% of the announced 600-person Superintelligence Labs reduction. Post-cut, SI Labs sits at roughly 3,000 people, meaning the total cut is about 17% of the division and the Menlo Park slice alone is about 10.6% of SI Labs headcount.

Why 318 is a bigger number than it looks

318 sounds small next to a Meta layoff cycle, but against the actual addressable market of Meta-trained AI research talent it is a market-moving event. In Refolk's index of professional profiles, only 148 U.S.-based Research Scientists and Research Engineers carry "Meta AI" in their headline, and the top current-employer bucket for that group is Meta itself with 23. The WARN pool is not 318 researchers, but the researcher slice of it is a meaningful fraction of every ex-Meta-AI person you could plausibly reach.

148

Compare that to where those people can land. Refolk's index shows 1,662 U.S. Members of Technical Staff and Research Scientists tagged to OpenAI, with roughly 60% of the sample concentrated in San Francisco and the Bay. OpenAI alone has about 11x the Bay Area research headcount to absorb Meta refugees with zero relocation friction. DeepMind Mountain View, Anthropic SF, and the Thinking Machines / SSI / Reflection cluster round out the demand side.

The full dataset

The mechanism is simple: rivals who move in week one exhaust the top decile fast. By week three the remaining names are either internally redeployed into TBD Lab, quietly negotiating with the labs that got there first, or intentionally sitting on severance until Q1.

The real competitor is TBD Lab, not Anthropic

If you are pitching a laid-off FAIR researcher this month, your competition is not another lab. It is Meta's own internal transfer desk. Reporting on the cut is explicit that Meta is encouraging affected employees to apply for other roles within the company, aiming to retain most of them, and the obvious internal landing pad is TBD Lab, the elite foundation-model team that was pointedly spared.

That reframes the pitch. You are not asking someone to leave Meta. You are asking someone who was just told, in effect, that they were not chosen for TBD to consider whether the internal transfer they can probably land is actually the offer they want. The frame that works:

• At TBD Lab, you are an also-ran to a nine-figure external hire
• At Meta broadly, you just watched your VP get cut
• At a rival lab, you are the hire, with named projects and compute you can point to

The named nine-figure hires the internal transfer would sit under: former GitHub CEO Nat Friedman, Safe Superintelligence co-founder Daniel Gross, Apple's former AI lead Ruoming Pang, and Thinking Machines co-founder Andrew Tulloch. That is a lot of gravity to compete with as an internal transfer, and a very easy story to tell as an outside recruiter.

You are not asking them to leave Meta. You are asking whether the internal transfer they can land is the offer they actually want.

Severance is leverage, not urgency

Do not run this as a fire sale. Meta's prior severance floor was 16 weeks of base pay plus two weeks per year of service, which for a typical FAIR tenure of five-plus years works out to roughly 26 weeks of runway. Recruiters who treat November as the deadline will get outbid in January by labs that used the extra 60 days to build a real research pitch.

What actually moves senior researchers on that timeline:

1. Compute guarantees in writing. A specific GPU allocation, not "we have capacity."
2. Research autonomy. Named projects they own, publication rights, conference travel budget.
3. Team continuity. The ability to bring one or two collaborators from the same WARN list.
4. A VP anchor. A principal or director they respect who is already committed.
5. Location fit. Menlo Park to SF is a commute. Menlo Park to London is a life event.

That last point is where the geography matters. The 60% Bay Area concentration in Refolk's OpenAI MTS sample tells you the frictionless offer is a same-metro one. DeepMind Mountain View is a short drive from 1 Meta Way. Anthropic SF is roughly 40 minutes. A London or Zurich pitch has to overcome a real move.

The three VPs are the whole game

The single highest-leverage move on this list is landing one of the three VPs, because each one can attract a team of 10 to 30 ICs from the same WARN roster within a quarter. The SFGate reporting names two software engineering VPs and one AI research VP without naming the individuals, but they are trivially identifiable from LinkedIn tenure filters plus the org context in the memo.

Two different playbooks here:

• Startups (Thinking Machines, SSI, Reflection, Physical Intelligence, Figure): chase the VPs. Offer equity, title, and a mandate. The IC pull-through is the actual prize.
• Big labs (OpenAI, DeepMind, Anthropic): chase the ICs the VPs will bring, but do it before the VP signs somewhere, because the moment a VP commits publicly, their former reports go into a two-week negotiation blackout with that destination.

This is where a plain-English search beats a Boolean string. Describing the person you want, rather than hand-tuning a query, is the kind of ask Refolk is built for: get a ranked shortlist across GitHub, LinkedIn, and the open web without stitching three tools together.

The robotics slice is the hidden arbitrage

The most under-priced group on this WARN is robotics, because Meta shed a robotics org at the exact moment Figure, 1X, Physical Intelligence, and Skild are scaling. Robotics IC talent with production experience is scarce industry-wide, and a Menlo Park robotics engineer who just got cut has four warm inbound destinations waiting if they know where to look.

For recruiters at humanoid and manipulation startups, the shape of the outreach is different from the researcher pitch. Robotics people move for:

• Real hardware in the building, not sim-only work
• A tight loop between perception, control, and deployment
• Equity math that reflects the current humanoid funding cycle

The GitHub activity and paper history are what tell you who was actually hands-on versus who managed, and that signal is what should drive the shortlist.

The 30-day sourcing plan

Here is the concrete week-by-week for the next month, assuming you are working the WARN pool cold. The Nov 21 badge-off date is your anchor, not the Dec 22 EDD date.

Week 1 (now through Nov 7)

• Pull the 318 WARN names against LinkedIn, GitHub, and Google Scholar. Tag each with team (search, voice, personalization, assistants, robotics, infra), tenure, and public output.
• Identify the three VPs by cross-referencing "Vice President, Engineering" and "Vice President, AI Research" against Menlo Park tenure of five-plus years.
• Prioritize the 148-headline researcher pool first, then robotics ICs, then infra.

Week 2 (Nov 8 to Nov 14)

• Reach out on work email while badges still work. This is the only week that channel exists.
• Lead with named team continuity: "I know two of your teammates on personalization. Would it help to talk about landing together?"
• For VPs, offer a specific mandate, not a title bump.

Week 3 (Nov 15 to Nov 21)

• Move to personal email and phone. Anyone who has not responded on LinkedIn by now is either negotiating internally with TBD or has already committed.
• Push VPs for a decision this week so their IC network is still coherent.

Week 4 (Nov 22 onward)

• Post-cliff. Focus shifts to the 26-week severance window and Q1 close.
• Build the compute and autonomy pitch in writing. This is what wins in January.

The tooling matters less than the discipline. 318 names against a 148-person addressable research pool means the top decile is gone by mid-November. Move accordingly.

FAQ

Which date should I plan around, Nov 21 or Dec 22?

Plan around Nov 21 for outreach and Dec 22 for negotiation. The SFGate reporting cites Nov 21 as the termination date, which is when access to internal systems and work email typically ends and the pool becomes reachable only through personal channels. The Dec 22 date in the EDD filing is when payroll formally stops and the severance runway begins. If you want work-email reach, you have until roughly Nov 20. If you want to close before Meta's internal transfer desk locks in TBD Lab redeployments, close before end of December.

How many of the 318 are actually AI researchers versus software engineers?

The WARN filing does not publish role breakdowns, but the reporting is explicit that dozens are software engineers and the list also spans data scientists, privacy program managers, and AI researchers across Superintelligence Labs, FAIR, and product infrastructure. Refolk's index shows only 148 U.S. profiles headline "Meta AI" as Research Scientist or Research Engineer, which puts a hard ceiling on how many of the 318 are pure research. Assume the majority are SWE and infra, with a smaller but disproportionately valuable research and robotics slice.

Is it worth pitching Meta researchers who might get TBD Lab offers instead?

Yes, and the pitch is easier than it looks. TBD Lab is small and elite by design, and the people on this WARN list are, by definition, the ones not selected for it. The framing that works is that an internal transfer into TBD makes them a supporting cast member to nine-figure external hires like Nat Friedman, Daniel Gross, Ruoming Pang, and Andrew Tulloch, whereas a rival lab offer makes them the hire. Compute, autonomy, and a named project beat title-and-comp against that backdrop.

What is the fastest way to find the three VPs on the list?

Filter LinkedIn for current or former Meta employees in Menlo Park with the titles "Vice President, Engineering" or "Vice President, AI Research," Meta tenure of five-plus years, and recent October or November 2026 activity. Cross-reference against the team areas SFGate named: search, voice, personalization, assistants, robotics. Three names will fall out.