SpaceX Just Bought Cursor for $60B. The Q3 Anysphere Poach Window Opens Now.

On June 16, 2026, SpaceX signed a $60B all-stock merger with Anysphere, the largest venture-backed startup acquisition on record. Close is expected Q3 2026. That gives founders and talent leaders roughly one quarter to build a live shortlist of the ~250 engineers who shipped the highest-revenue AI coding tool of the cycle, before every AI infra recruiter in San Francisco is doing the same thing.

Why the Anysphere roster is the tightest poach window of 2026

Because every vested engineer at Cursor is about to hit maximum liquidity and minimum culture fit in the same week. The deal converts all outstanding Anysphere shares to SpaceX Class A stock priced on a seven-day volume-weighted average before close. Newly liquid public paper, a 5-day in-office SF mandate, and a new reporting line into Elon Musk's stack: that is the exact combination that ends tenure at MIT-dorm startups.

The cofounders already sketched the template. Arvid Lunnemark, 26, left in October 2025 to start Integrous Research on "safer AI" - a direct thematic rebuke to joining xAI. He did it a full year before the deal was public. Senior ICs who chose Cursor specifically for model-agnosticism (Cursor currently routes queries through Claude and GPT) are about to report into a competitor to both. Ideologically motivated engineers do not wait for the vest cliff.

$13.3M

What actually changes at close

At close, three things flip at once for every Anysphere IC: equity becomes liquid, culture becomes SpaceX, and the model layer becomes captive to xAI's Grok. Any one of those triggers is enough to open a conversation. Together they define the shortest, richest sourcing window in AI coding since Copilot shipped.

Here is the specific mechanic behind each trigger:

• Liquidity. All Anysphere shares convert to SpaceX Class A on a 7-day VWAP. Employees who have been paper-rich for 18 months finally have something they can borrow against or sell into secondary. The golden-handcuff argument weakens the day the ticker prints.
• Culture. Cursor already runs a strict 5-day in-office policy at its SF HQ. Layer SpaceX-style intensity on top and you lose the one thing $60B in stock cannot replace: startup optionality and geographic flexibility.
• Model politics. SpaceX buying Cursor lets xAI redirect developer mindshare toward Grok. Engineers who built the model-agnostic router now ship features that route away from Anthropic and OpenAI. That is not a neutral change for the people who wrote the router.
• Regulatory overhang. SpaceX owes $10B if the deal collapses and $4B if antitrust blocks it. Those are unusually large reverse-termination fees. Anyone with a Q3 or Q4 vest cliff has real incentive to negotiate double-triggers now and start taking pitches this month.

The four cohorts you are actually sourcing

Anysphere is not one talent pool. It is five stacked cohorts with different exit incentives, and treating them as one is why generic outreach will fail.

1. The original Cursor engineering core. ~150 people who joined when Anysphere was still an MIT dorm project. Most equity, most identity attached to model-agnosticism, most likely to found rather than jump.
2. The Supermaven cohort. Folded in when the standalone product wound down in late 2025. Already survived one integration. They know the second one is worse.
3. The Koala cohort. Anysphere hired top Koala engineering talent in July 2025 as an acqui-hire. Small group, high signal, likely on retention grants that were pre-priced against Anysphere equity, not SpaceX.
4. The Graphite cohort. Anysphere agreed to acquire Graphite in December 2025. Newest arrivals, weakest ties to the Cursor mission, most likely to interpret the SpaceX deal as "the company I joined does not exist anymore."
5. Named senior hires. Travis McPeak (ex-Resourcely CEO) came in to lead security. Founder DNA, obvious double-trigger candidate.

Each cohort responds to a different pitch. The Supermaven and Graphite cohorts are looking at their third employer inside 18 months and want stability. The original core wants either seed money or a new hard problem. Sending the same InMail to both is why generic recruiter response rates on this pool will collapse within four weeks.

The market share numbers that reframe the pitch

Cursor's enterprise spend share fell from ~41% in June 2025 to ~26% in May 2026 according to Ramp, while Menlo Ventures puts Claude Code at 54% of the AI coding market. Both numbers are true. They measure different things, and that gap is your opening line.

Ramp and Menlo disagree by roughly 4x on Anthropic's share because one tracks credit card spend and the other surveys enterprise buyers. The practical implication for sourcing: Cursor engineers still believe they are winning on enterprise revenue (they are, on Ramp's methodology) but they also know Claude Code is winning on mindshare. That cognitive dissonance is the wedge.

The wrong recruiter frames the move as "before the ship sinks." The right one frames it as "before the SpaceX overhang caps your upside on a company that is still doing $13M ARR per head."

The engineers who built a model-agnostic router now report into a frontier-model competitor. That is the pitch, not the price.

How to build the shortlist before Q3

Start with named cohorts, not job titles. The Anysphere org chart is public enough via LinkedIn, GitHub commit history, and prior-company overlap that you can rebuild it in a week if you know what you are looking for. What you are looking for is: current Anysphere tenure, prior employer (Supermaven, Koala, Graphite, or original), public GitHub activity on the Cursor extension repos, and any signal of a personal blog or Twitter presence on model routing, evals, or IDE tooling.

That last filter matters more than titles. The ICs worth chasing are the ones who published something opinionated about tab completion, speculative decoding, or eval harnesses in the last 18 months. Those are the people other AI coding startups will pay $1M+ to move, and they are the ones who will not tolerate a Grok mandate.

This is the exact gap Refolk closes for a sourcing team on a Q3 deadline: you describe the person in plain English ("engineers at Anysphere who joined via the Supermaven acqui-hire and have public writing on model routing") and get a ranked shortlist across GitHub, LinkedIn, and the open web. No 40-line boolean, no scraping LinkedIn Recruiter for a week.

The compensation floor to plan against

Cursor engineer total comp reportedly runs $808K to $1.28M+, with a median around $1.1M/year, heavily weighted toward equity. That is the number a poacher has to beat, and the equity component is the piece that becomes negotiable at close: SpaceX stock is liquid but not appreciating at Anysphere's growth rate. A seed-stage AI infra startup offering 1-2% and a technical mandate can compete on expected value in a way a $200K base bump cannot.

$200M

The five landing pads to expect competition from

Every Anysphere engineer worth hiring will have offers from the same five places by August 2026. Anthropic, OpenAI, and three categories of challenger.

• Anthropic. Direct model competitor, Claude Code owner, ideologically opposite pole from xAI. The obvious first offer for anyone who left Cursor over model politics.
• OpenAI. Codex team is rebuilding after Cursor ate their category. Warm intros through investor overlap and the Sam Altman network.
• Windsurf / Codeium / Z.ai and other stealth AI-coding startups. Direct product competitors, strongest pitch on "keep building the model-agnostic version."
• Menlo Ventures portfolio. Menlo published the market share data; Menlo also has portfolio companies that need exactly this profile. Warm intros are already routed.
• New founding teams. The Lunnemark template. Expect 5-10 new AI coding or dev tools startups to spin out of Anysphere in the first year post-close, most seeded before the deal formally closes.

If you are an in-house talent leader, you are competing with all five and with the founders' own recruiting networks. Speed of shortlist matters more than volume of outreach. Getting to the right 30 people in week one of Q3 beats reaching out to 300 people in month two, which is another place a natural-language query into Refolk beats a boolean-then-enrich pipeline: you get to the right 30 the same day you decide to look.

What Q3 looks like day by day

Assume close in mid-Q3. Here is what a disciplined sourcing calendar looks like.

• Weeks -8 to -4 (now). Build named shortlist of ~150 targets across the five cohorts. Layer public writing, GitHub, and prior-company signal. Do not reach out yet.
• Weeks -4 to -1. Warm intros only. Founders and technical hiring managers reach out to the top 30 through mutual connections. Frame as "long conversation, no urgency."
• Week 0 (close). Public announcement of vested stock convert. Send direct outreach to the full 150, referencing the specific work you know they did (extension architecture, eval harness, RLHF pipeline for tab).
• Weeks +1 to +4. First wave of departures becomes public on LinkedIn. Follow-up on the remaining 120 with "we saw X leave, we know what you are weighing."
• Weeks +4 to +12. Second wave. This is when the Supermaven and Graphite cohorts move in size.

The teams that already have this calendar drafted will get 3-5 of the 30 they wanted. The teams that start when they see the LinkedIn banners flip get whoever is left over, which by month three is nobody worth $1M.

FAQ

When exactly does the SpaceX Anysphere deal close?

Q3 2026, per the June 16 announcement. All outstanding Anysphere shares convert to SpaceX Class A stock priced on a seven-day volume-weighted average of SpaceX's share price immediately before close. If the deal collapses SpaceX owes $10B, and $4B if antitrust blocks it, so a delayed close is possible but a fully abandoned close is unlikely at those fee levels.

How many engineers are actually at Anysphere?

Public sources put Cursor by Anysphere at approximately 300 people as of 2026, with engineering-only estimates around 250. That includes at least three prior acqui-hire cohorts (Supermaven, Koala, and Graphite, the last still integrating from the December 2025 deal) plus named senior hires like Travis McPeak on security. Treat those cohorts as separate pools with separate pitches.

What is the realistic compensation to beat?

Reported total comp runs from $808K to over $1.28M, with a median near $1.1M/year and heavy equity weighting. The equity piece is the negotiable part at close, because SpaceX stock is liquid but not growing at Anysphere's ~$13.3M ARR per employee rate. A seed or Series A AI infra startup offering meaningful founding equity and a technical mandate can compete on expected value even if the cash number is lower.

Where should I start if I only have time to run one query this week?

Start with the intersection of Anysphere current employment, prior-company signal (Supermaven, Koala, or Graphite), and public writing on model routing or IDE tooling in the last 12 months. That is roughly 30 to 60 people, they are the highest-signal targets, and every other AI coding recruiter will find them by August. Running that as one plain-English query into Refolk gets you the ranked list the same afternoon, which is the whole point of building the shortlist before the announcement banners flip.