All 11 xAI Co-Founders Are Gone. Anthropic Already Got Nordeen.
xAI lost every original co-founder and at least nine engineers by March 2026. Here is how to source the cohort before Anthropic and Meta MSL close them out.
Between February 10 and late March 2026, xAI lost every single one of Elon Musk's original 11 co-founders, plus at least eight more senior engineers in a single week. Musk called it a "reorganization to improve speed of execution." For recruiters, it is something simpler: the most concentrated frontier AI free-agent pool of the year, and one that is largely invisible on LinkedIn.
If you wait for "Open to Work" badges, you will lose. Ross Nordeen, Musk's last remaining co-founder, was at Anthropic within roughly 60 days of leaving. Anthropic announced his hire the same day it announced a deal to use xAI's Colossus supercomputer. That is the template competitors are running. The window on this cohort is measured in weeks.
What actually happened at xAI
The timeline is tight and worth getting right, because the dates drive your outreach windows.
On February 2, 2026, SpaceX acquired xAI in an all-stock deal valuing the company at $250 billion. That detail matters: this cohort is sitting on liquid equity and has no meaningful comp lock-in pulling them back.
The week of February 10, TechCrunch reported that two more co-founders had exited, bringing the total to six of the original 12, and that at least nine engineers including those two co-founders had announced exits in a single week. NBC News later confirmed Musk's own framing on X: the company was "reorganized" and the process "required parting ways with some people."
By March 28, the last two co-founders, Manuel Kroiss and Ross Nordeen, were out. That is the full set. All 11 are gone.
Two weeks earlier, on March 12, Andrew Milich and Jason Ginsberg announced they were leaving Cursor to join xAI and SpaceX, both reporting directly to Musk. At Cursor, the two oversaw product engineering that took the company to a $2 billion annualized revenue run rate. That hire is the tell on what xAI is rebuilding: product engineering, not research. Which means the departed research cohort has no natural path back into xAI. They are permanently in the market.
Why this is the buy of June
Forget the abstract "frontier AI talent" framing for a second. Look at who actually left.
Jimmy Ba co-authored the 2014 Adam optimization paper. It is the most-cited paper in modern AI, with more than 95,000 citations. He is, by any reasonable measure, the highest-prestige free agent in the market right now.
Tony Wu ran reasoning. Kyle Kosic came from OpenAI. Greg Yang came from Microsoft Research. Toby Pohlen spent six years at Google before xAI. Zihang Dai and Guodong Zhang round out the research side. On the engineering side, Vahid Kazemi worked on multimodal and told the world publicly that "all AI labs are building the exact same thing, and it's boring," before saying he was starting something new. Shayan Salehian spent 7+ years across Twitter, X, and xAI on product infrastructure. Hang Gao worked multimodal on Grok Imagine.
There are three clean buyer profiles here, and it helps to source against all three in parallel rather than treat the cohort as a single bucket:
- Researchers (Ba, Wu, Dai, Zhang, Yang). Going to Anthropic, Thinking Machines, or staying stealth on a paper trail.
- Infra and program operators (Nordeen, Pohlen, Kroiss). Going to Anthropic compute, Meta Superintelligence Labs compute orgs, or back into hyperscaler infra teams.
- Product engineers (Salehian, Kazemi, Gao). Going to Cursor, smaller labs, or founding new companies.
By July, the named cohort is gone. What you can still catch in June is the long tail of senior ICs who never tweeted their exit.
Why LinkedIn alone will fail you here
Here is the operational problem. Most of this cohort announced their departure on X, not on LinkedIn. NBC reported that xAI did not respond to requests to verify whether the people behind several of those X posts had actually worked for the company. A pull on standard professional-network data for xAI Grok researchers in the US returns essentially no current public matches. The cohort is not tagged. The cohort did not update their headlines. Many of them never had public LinkedIn profiles with "xAI" on them in the first place, because their entire careers ran inside Musk's orbit (Salehian is the archetype here: Twitter, then X, then xAI, with no clean external paper trail).
If you build a LinkedIn Recruiter search on "xAI" + "past company" + "left in last 90 days," you will get a near-empty pond. The signal is not on LinkedIn. The signal is spread across:
- X bio changes and pinned posts (the primary departure signal for this cohort)
- GitHub commit history dropping off on xAI-affiliated repos
- arXiv co-authorship on Grok and related papers
- Conference talk pages (NeurIPS, ICML, ICLR speaker bios)
- Patent filings under xAI assignment
- Personal blog posts and Substack starts
This is exactly the multi-source sourcing problem we built Refolk to handle. You describe the person in plain English ("ex-xAI research scientist who co-authored a Grok paper, left after January 2026, based in the Bay Area") and get a ranked shortlist pulled from GitHub, X, paper authorship, and the open web in addition to LinkedIn. LinkedIn alone, for this cohort, is fishing where the fish aren't.
The comp math actually favors you
A common reaction to "ex-xAI researcher" is "I cannot compete with Meta MSL." Reasonable instinct, wrong conclusion.
Meta has reportedly offered packages worth up to $300 million over four years to retain top AI researchers. Andrew Tulloch joined Meta Superintelligence Labs in late 2025 with a deal the Wall Street Journal reported as roughly $1.5 billion over six years. Those numbers are real.
But they apply to a handful of named people at the top of the prestige stack. For the rest of the cohort, the comp picture is different. Per Levels.fyi, OpenAI L5 software engineers sit at around $1.15 million total comp. Anthropic median total comp is around $600K. xAI 2026 base pay sits at $180K to $440K in Seattle and the Bay Area, putting it mid-frontier tier, below OpenAI and Anthropic.
Translation: most of the departed xAI engineers, the ones who are not Ba or Wu, can be moved without nine-figure offers. You do not need to outbid Meta MSL. You need to out-pace Meta MSL's outreach. That is a tractable problem if you start now.
A practical 14-day plan
Stop reading and run this.
Day 1 to 3: Build the named list
Start with the 11 confirmed co-founders. Add the named engineers from the reporting: Kazemi, Salehian, Gao, plus anyone else flagged in the TechCrunch and NBC coverage. For each, find their X handle, their GitHub, and their most recent paper. This is the seed set.
Day 3 to 7: Expand to the long tail
This is where most recruiters stop, and where the real opportunity sits. The named cohort is being chased by everyone. The long tail of senior ICs who left quietly during the February reorganization is the actual edge. Pull every co-author from xAI papers between September 2024 (Series B) and February 2026 (SpaceX acquisition). Cross-reference against GitHub commit drop-offs on xAI-affiliated repos. Cross-reference against X bio changes that dropped "xAI" in the last 90 days.
A natural-language query against a multi-source index is the fastest way through this step. The structured filters on LinkedIn Recruiter will not get you there, because the underlying data is not on LinkedIn. This is the part of sourcing AI engineers in 2026 that has genuinely changed: the signal lives on GitHub and X first, and the platform you used in 2022 cannot see it.
Day 7 to 10: Triage by buyer profile
Split your list into the three buckets above (research, infra, product). Different pitches, different timelines. Researchers care about compute access and co-author quality. Infra people care about scope and reporting line. Product engineers care about shipping velocity and equity upside, which is why Cursor is winning this segment.
Day 10 to 14: Outreach, in that order
Researchers first, because Anthropic is already in motion. Infra second, because Meta MSL compute is the next likely move. Product engineers last, because that segment has the longest decision cycle and the founders among them (Kazemi is the archetype) want to talk to investors before recruiters.
If you are operating solo or on a small team, this is roughly a week of work with a tool that can search across GitHub, LinkedIn, and the open web in one query. We built Refolk for exactly this shape of problem: a recent, news-driven cohort that does not announce itself on LinkedIn.
The bigger pattern this fits
This is not a one-off. Helion, Oracle, Bungie, Stainless: every month in 2026 has produced a concentrated, news-driven cohort that resists standard sourcing. The xAI reorganization is just the cleanest example because the names are public and the prestige stack is unusually deep.
The recruiters who are winning these windows have stopped treating LinkedIn as the index and started treating it as one source among several. The xAI cohort is a particularly stark case (almost zero current LinkedIn signal), but the same playbook applies to the next reorganization, and the one after that.
The xAI layoffs of 2026 are a sourcing event, not a news event. Treat them accordingly.
FAQ
How many people actually left xAI in the reorganization?
Public reporting confirms all 11 of Musk's original co-founders, plus at least eight additional senior engineers exiting in a single week in February 2026. The actual count is almost certainly higher because most exits were announced on X, not LinkedIn, and xAI declined to verify many of them to NBC. Treat 20+ as a floor for named departures, with a long tail of unnamed senior ICs underneath that.
Why is Anthropic winning this cohort so fast?
Anthropic is pairing compute deals with talent absorption. Ross Nordeen's hire was announced the same day as Anthropic's deal to use xAI's Colossus supercomputer. That pattern (infra relationship as a talent funnel) is the template, and it explains why Anthropic is consistently the first competitor named when an xAI co-founder lands somewhere. If you are not Anthropic, you are not getting in through that door, which means you need to be faster on outreach.
Can I really compete with Meta Superintelligence Labs on comp?
For the top three or four names in the cohort, no. Tulloch-tier deals are nine figures and Meta is willing to pay them. For the other 90% of the cohort, yes. xAI base maxes near $440K, Anthropic median is $600K, and most of this cohort is sitting on already-liquid equity from the SpaceX acquisition. The constraint is outreach speed, not budget.
What is the single highest-leverage move this week?
Build a list of every paper co-author on xAI publications between September 2024 and February 2026, cross-reference against GitHub activity drop-off, and reach the long tail before anyone else does. The named co-founders are being chased by every recruiter at every lab. The 30 to 50 unnamed senior ICs underneath them are the actual buy, and they are reachable for another two to three weeks before competitor outreach catches up.