Mitchell Hashimoto wrote on April 28 that Ghostty is leaving GitHub:

Ghostty will be leaving GitHub.

I read that and had the annoying reaction of understanding both sides.

GitHub is still useful to me. I use it constantly. I still get a lot of value from it. Also, if I had spent years living in issues, PRs, Actions, releases, and random project pages, and that surface started failing during the hours I needed it, I would be mad too.

Then I went back to cleaning up pending PRs and GitHub showed me this:

GitHub warning banner saying some pull requests may be missing because of an ongoing search incident.

Come on.

The banner said some pull requests may be missing because of an ongoing search incident. GitHub’s status page called it “Incomplete pull request results in repositories.” It started on April 28 at 14:17 UTC. GitHub said the PR data was not lost, but pages relying on Elasticsearch might show incomplete results while reindexing continued. The workaround was the API or gh pr list.

I know the Git repo is distributed. That does not help much when the thing I am trying to do is review PRs and the PR page is the broken part.

Mitchell says in a footnote that the timing was coincidental with the big April 27 outage. They had been discussing the move for months, and the post was written more than a week earlier. Fair.

Still, the timing is brutal. A well-known open source maintainer says Ghostty is leaving GitHub, and while I am reading it GitHub is telling me my PR list might be incomplete.

I had just written about GitHub feeling the agentic coding load. That post was mostly capacity plans and queue pressure. This was the same problem showing up as: I wanted to review PRs and the PR list might be lying to me.

I think that is why Mitchell’s post landed for me. He is not someone casually dunking on GitHub. He wrote:

I’m GitHub user 1299, joined Feb 2008.

He also talks about reading GitHub issues on vacation and wanting to work there because of Vagrant. I laughed at that because I get it. Some people open social apps when they have a spare minute. I open project pages and issues and weird old repos. Normal, healthy behavior. I should probably not defend it too hard.

The user number made me curious, so I checked mine:

gh api users/ashokgelal

I am user 401055, created on September 15, 2010. I swear I had joined earlier, but apparently memory is not append-only. Maybe I was using Google Code. That would at least make my memory less dramatic.

Very rough math, using February 1, 2008 as Mitchell’s starting point because the post only says February 2008:

  • 401055 - 1299 = 399756 user IDs between us
  • about 31.5 months
  • about 12,700 IDs per month
  • about 417 IDs per day
  • about a 309x increase in the ID number

If you abuse that as compound growth, it is around 20 percent per month. This is not real GitHub growth analysis. User IDs are not active users. Mitchell’s exact join day would move the numbers. I just wanted a feel for the gap.

September 2010 still felt early at the time. February 2008 is another planet.

I am not moving off GitHub because of this. I do not maintain a high-stakes open source project at Ghostty’s scale. For my work, GitHub is still where I search first. If I need a library, a lot of my searches are basically:

site:github.com thing I need

I almost never think of GitLab unless someone else says it first.

So I am not predicting a GitHub exodus. I have no special data there, and one maintainer moving one project is not a trend.

But this one is worth noticing because it is not coming from someone who never liked GitHub. It is coming from someone who liked it too much and finally got practical.

That feels worse.