The painful death of Github
TL;DR
Theo says GitHub has crossed from annoying to untrustworthy — the breaking point wasn’t just outages, but incidents where merged pull requests appeared reverted, which he calls a split-brain failure that makes the merge button itself feel unsafe.
The reliability picture looks far worse than GitHub’s official status page suggests — he cites Mere’s “missing GitHub status page,” which counts any user-visible service disruption as downtime and puts GitHub at 86.75% uptime instead of the cleaner, service-scoped numbers on the official dashboard.
The response from GitHub leadership is a huge part of the backlash — Theo tears into COO Kyle’s post for describing reverted merges as a “regression in merge queue behavior” affecting 2,840 PRs out of 4 million, arguing the language was crafted to minimize one of GitHub’s most serious failures ever.
His bigger diagnosis is organizational, not just technical — Theo argues GitHub effectively has no CEO, now reporting into Microsoft’s CoreAI org, and says a hard split between product and engineering only works with strong top-level leadership, which he believes is gone.
This hit especially hard because GitHub was personal for a generation of open-source developers — he echoes Mitchell Hashimoto, GitHub user #1299, who said he cried writing Ghostty’s announcement that it’s leaving GitHub after 18 years of daily use and weeks of outages disrupting serious work.
Theo ends by saying GitHub is now failing all four trust layers — does it work like before, does it work right now, does your work persist, and can others exploit the platform to harm users, pointing to a same-day GitHub RCE bug and Tanner Linsley’s unresolved npm/TanStack squatting mess as proof.
The Breakdown
GitHub, but make it heartbreaking
Theo opens less like a pundit and more like someone grieving. He says GitHub shaped his career, friendships, and identity as a developer over 15 years, which is exactly why watching it “die” feels so personal instead of like just another SaaS complaint.
The outages stopped being annoying and started feeling dangerous
The immediate trigger was a day when he couldn’t even load pull requests for T3 Code, where they have hundreds open. But the bigger alarm bell was a separate incident where already-merged changes appeared reverted — the kind of split-brain state that can leave production, webhooks, and repo history disagreeing with each other in ways that are miserable to debug.
Why the uptime numbers feel fake to users
Theo walks through GitHub’s official monthly uptime chart, then says it undersells what developers are actually living through. He points to Mere’s alternative status page, which treats any partial outage as downtime, and says that framing drops GitHub to 86.75% uptime — “there’s no nines” — which matches the lived experience much better.
The moment his trust first cracked: a 2022 webhook outage
He tells a story from VidCon 2022: he fixed a bespoke iPad app issue for an important client, merged the code, and expected Vercel to auto-deploy. Instead, GitHub webhooks were down, some events sent, some queued, some vanished, and he had to rip out the integration and deploy manually from his laptop just to get the fix live.
Mitchell Hashimoto’s exit made the whole thing feel real
Theo reads from Mitchell Hashimoto’s post announcing Ghostty is leaving GitHub, including the line that he cried while writing it. Mitchell, GitHub user #1299 since February 2008, says he opened GitHub every day for 18 years and had started marking an X in a journal every day an outage blocked his work — which lately was “almost every single day.”
The merge-revert post and the leadership vacuum
Theo absolutely unloads on GitHub COO Kyle’s response, especially the phrasing: “regression in merge queue behavior,” “appear reverted,” and “0.07%” of PRs affected. To him, that language reveals a deeper problem: GitHub has no CEO, reports into Microsoft’s CoreAI org, and lacks the kind of accountable leader who can own failures instead of minimizing them.
A broken org chart, not just broken systems
From there he zooms out to structure. He says GitHub has a hard wall between product and engineering — which he sees as especially absurd for a developer tool where the users are engineers — and argues that setup only works if both sides report into a strong leader who can keep them aligned; without that, it becomes what he bluntly calls “a dead company.”
Security failures and npm neglect pile on
Right before filming, Theo says a Wiz researcher found an RCE bug in GitHub that could access millions of repositories via unsanitized push options embedded in internal headers. He also brings up Tanner Linsley’s long-running fight over the squatted TanStack npm package, saying GitHub/npm ignored obvious warning signs until a malicious package started exfiltrating .env files — proof, in his view, that GitHub is no longer merely inconvenient but actively failing maintainers and users.
Theo’s four levels of trust — and why he thinks GitHub failed all of them
He ends by formalizing the argument: reliability means the product behaves the same way, works right now, preserves yesterday’s work, and protects users from malicious abuse. In Theo’s view, GitHub is now failing every one of those tests, which is why he says he can no longer recommend it and is going to start evaluating alternatives aggressively.