SpaceX Bought Cursor for $60 Billion. Days Later, It Launched a GitHub Rival.

SpaceX closed its $60 billion buyout of Cursor's maker on Aug 14. Days later, Cursor launched a GitHub rival the day GitHub had a 7h47m outage.

SpaceX closed a roughly $60 billion all-stock acquisition of Anysphere, the company behind the Cursor code editor, on August 14, 2026. Three days later, on August 17, Cursor launched Origin, a code hosting platform built to do what GitHub does: repositories, pull requests, code review, a place your commits live. GitHub spent that same day offline, a capacity failure that ran nearly eight hours. The overlap made for an easy headline. The real story for an engineering team is not the coincidence. It is that a code editor's parent company and a git host's core business just became two very large, very different bets, neither of which is about hosting code well for its own sake.

The acquisition is the more startling number on its own. SpaceX disclosed the deal in a Form 8-K filed with the SEC on June 16, 2026, exercising an option it had held since April to either pay Anysphere roughly $10 billion for a partnership or acquire the company outright for $60 billion later in the year. It chose the second path. Coverage at the time called it the largest acquisition of a venture-backed startup in history. The deal closed August 14, all stock, and folded Cursor into a business built around rockets, satellites, and Starlink.

Sources: SpaceX Buys Cursor In Largest Startup Acquisition Ever At $60 Billion (Forbes), Cursor capitalizes on GitHub frustration, launches rival hosting platform (TechCrunch)

Origin covers the core of what GitHub does. It hosts repositories, which Cursor's own changelog treats as the authoritative source once a project lives there. It handles pull requests with a full timeline, commits, checks, and diffs. It syncs two directions with existing GitHub repos, so a comment posted inside Cursor shows up on GitHub within seconds and the reverse also holds, which means a team can point Origin at a project without migrating it outright. Three integration partners shipped on day one: Vercel for pull request preview deployments, and Depot and Buildkite to run existing GitHub Actions workflows or native pipelines. None of this appeared from nothing. In December 2025, Cursor acquired Graphite, a code review platform already used by hundreds of thousands of engineers, and its workflow became the review layer inside Origin.

Sources: Origin: code hosting for the AI era (Cursor changelog), Graphite is joining Cursor (Cursor)

This post argues that a coding tool and a git host should be evaluated as one dependency decision instead of two, and that argument favors a category this piece comes from: we build TLM Forge, a review and process layer meant to sit on top of whatever agent, model, or host a team already runs, specifically so a change at any one vendor does not change how code gets reviewed. Read the vendor-neutral parts of this post as the load-bearing claims, and read the one product section knowing where it comes from.

What Actually Changed, and What Didn't

GitHub is not going anywhere soon. Its 2025 Octoverse report puts the platform at more than 180 million developers, a base Origin's rollout to paid Cursor plans does not come close to touching. Origin does not force anyone off GitHub either. The two-way sync exists specifically so a team can try Origin on one project while GitHub keeps serving everything else, and enterprise organizations can opt out of the rollout entirely. What changed is narrower and more structural: GitHub's most credible new competitor is no longer an independent, developer-tools-only company. It is a feature inside the same editor millions of engineers already run their AI agents in, owned by a company whose primary business is orbital rockets and satellite internet.

Sources: Octoverse 2025: 180 million-plus developers (GitHub Blog)

That structure matters more than the feature list. A git host used to sit closer to a shared utility: stable, boring, nobody's strategic centerpiece. GitHub was already not neutral in the purest sense, Microsoft has owned it since a $7.5 billion all-stock deal closed in 2018, but Microsoft's core business is still software and cloud infrastructure, adjacent to what GitHub does all day. SpaceX's core business is not. When a code host becomes a line item inside a much larger, unrelated bet, its roadmap, its uptime priorities, and its risk tolerance answer to that bet first, not to the engineers who happen to depend on it daily.

Sources: Microsoft to acquire GitHub for $7.5 billion (Microsoft)

The Outage Was a Capacity Failure, Not an AI-Written Bug

It would make a tidy story if AI-generated code had taken GitHub down on the exact day an AI coding company launched a competing host. That is not what happened, and GitHub said so directly in its own postmortem: "Neither outage was caused by a code or configuration change. Both incidents were capacity failures at their core," referring to the August 17 outage alongside a separate GitHub Actions failure on August 6. The August 17 incident ran from 13:28 to 21:15 UTC, seven hours and 47 minutes, with error rates of roughly 20 percent across the website and API and roughly 50 percent on archive and raw file downloads. The trigger was a critical infrastructure component in GitHub's Central US data center that failed to scale when traffic hit a new peak, worsened by a client-side retry loop that added load while GitHub was already trying to recover.

Sources: The August 17 outage, and the work ahead (GitHub Blog), GitHub Status: incident history

The scale behind that traffic peak is real and disclosed in the same postmortem: monthly commits on GitHub roughly doubled, from 1.4 billion in April 2026 to 2.9 billion by August. GitHub does not break that growth down by source, but it lines up with an industry-wide shift toward agent-generated commits rather than commits typed by hand. Capacity planning built for one growth rate met a much steeper one, and a single data center's infrastructure did not scale fast enough to absorb it. That is an operations failure, not a code quality failure, and the distinction changes what a team should actually take from it.

Sources: The August 17 outage, and the work ahead (GitHub Blog)

The Real Question: What Did You Just Outsource

Neither event is a reason to migrate off GitHub or to sign up for Origin's rollout. It is a reason to write down what a team actually depends on, because the honest answer changed twice in one week without anyone on a typical engineering team casting a vote. Four questions matter more than the feature comparison before either platform becomes load-bearing for daily work.

1. Whose roadmap does your host actually serve

Ask what the parent company optimizes for, because that answer eventually becomes the host's answer too. A pure-play developer tools company survives by keeping the engineers in front of it satisfied with the product itself. A company where git hosting is a strategic asset inside a much larger bet, whatever SpaceX ultimately decides Cursor is for internally, optimizes for that bet first. Neither answer is automatically disqualifying. The mistake is not asking the question until a host is already too embedded to leave.

2. Can you actually get your history out

Test the exit, not the pitch. Origin's two-way GitHub sync is a genuine advantage here, since a team can point it at existing repositories without a hard cutover. But sync that behaves cleanly during a calm pilot is not the same guarantee as an export that behaves cleanly under pressure, mid-outage, or after a company the size of SpaceX decides Cursor's hosting priorities should shift. GitHub's own export and migration tooling has had years of real load to get boring. Origin's has had weeks.

3. What happens when the reviewer and the host share an owner

Origin puts the code, the pull requests, and the AI agent that wrote the change in one interface from one vendor. That is a real workflow win and a real governance question at the same time: when the tool that writes the code, the tool that hosts it, and increasingly the tool that reviews it all come from the same company, who is checking that reviewer's incentives against the team's? Atomic commits that isolate one logical change stay reviewable and portable no matter which platform ends up holding them, which is exactly the property worth having while the platform question itself is unsettled.

4. What is your blast radius when it goes down

An outage that used to sideline one tool now risks sidelining several at once, because bundling an editor, an agent, code hosting, and review into a single vendor's stack means a single vendor's bad day reaches further into daily work. GitHub's August 17 outage cost roughly eight hours across the website, API, Actions, and Copilot together, all served from infrastructure with a shared failure point. Adopting a second, more tightly bundled platform without a rollout plan just relocates that concentration risk instead of removing it. A staged rollout with one review gate and CI guardrails that hold regardless of which vendor sits underneath them is the difference between adopting a new host on purpose and discovering the concentration risk during the next outage.

This is close to the problem the convergence gate is built to solve, applied to infrastructure rather than a single diff. The gate blocks a code change from shipping until independent review agents converge on zero critical issues, specifically so the tool proposing a change is never the only one checking it. The same principle applies to picking a git host or a coding platform: the vendor best positioned to sell a team on Origin is not the party best positioned to say when it is not ready for that team, and the same is true of GitHub insisting an eight-hour outage does not change anything. An independent check, whether it is a second reviewer, a second agent, or a rollout plan someone outside the vendor signed off on, is what keeps that decision honest.

The decision itself is also easy to lose. Why a team picked GitHub over Origin, what got tested before a pilot repository moved, and what the fallback plan was if a sync broke mid-migration are exactly the details that live in one engineer's memory and disappear when that engineer changes teams. That record belongs in a decision log the next engineer can actually search, not in a Slack thread that scrolls out of reach. MemX, from the same team behind TLM Forge, addresses an individual's version of the same durability problem: a private, persistent memory layer for personal photos, documents, voice notes, and messages, not team platform decisions.

GitHub is not down for good, and Origin is not yet a real threat to its 180 million developers. What changed this month is that two companies now treat code hosting as a strategic feature of a much larger business rather than a product that has to stand on its own, and a team that has not worked out what that means for its own dependency is behind on the question regardless of which host it happens to use today.