"Is GitHub dead?" Recorded a spontaneous little video with thoughts on the future of developer tooling

Right on, the sort of gift economy around code contributions has changed completely now that code is so much easier to write. IMO the idea of external/non-expert code contribution feels like an anachronism. I wrote on this here:

That's a very, very good post, thank you!

Here's a full transcript if that's more your thing:

I know that I should've cleaned my glasses btw

Great content, man thank you but it's short form video content. You can hold the phone in the vertical portrait orientation. That's how I'm watching it right now in the bathtub. I'm not watching it on my fucking home cinema

Love the video! Look forward to more of them. Maybe I'll even give one a go myself! I agree with some of the things you said. GitHub has some really human-centric features but it also provides cheap and long term storage, conflict resolution, "multiplayer" interactivity (both for humans and agents), audit history and the ability to roll-forward/back very quickly. As agents and humans write more and more code, I think these things become MORE valuable. You could argue that most of the things I mentioned come from git as a protocol and not Github specifically, which is a fair argument. Maybe there is room for something new to be built on the protocol that is more agent-centric but also has backwards compatibility to how we built up until now. However, GitHub/GitLab is well positioned to deliver on this, so I personally wouldn't bet against it being prominent in our future.

I agree with all of the things GitHub "also provides", but I don't think long-term it's a good strategy to bet on a product based on what it "also provides", i.e. if you told me that no one watches YouTube anymore but people use it to store their home videos, I wouldn't bet on YouTube. Basically: yes, the things you mentioned are nice, but not why I would bet on GitHub in the video. And also, yes, they *are* well positioned to take advantage here, but the question is whether they will or not. Things are moving extremely fucking fast nowadays and I don't see GitHub moving as fast as they should be.

I've coded things you people wouldn't believe. Handcrafted SAX parsers processing legacy SOAP requests on fire off the shoulder of US-WEST-1. I watched C-String implementations glitter in the dark near XÖRs Gate. All those moments will be lost in time, like tears in rain. Time to die.

my girlfriend asked whether you are sitting on a Traktor :D

No, a bench!

Also because, y’know… Microsoft. Same applies to VS Code. By the time innovation actually happens it’ll be too late

Code was never expensive. You could've outsource to a 3rd world country and get cheap labor. But as always, you got what you paid for. Same is true for LLMs.

love this, more of these please 💪

hot damn, okay okay

back to playing some factorio

Dont think so Github has more 3 4 years.

Ha ha ha….. 10 years This guy is sharing some golden nuggets, then shares his “10 years” estimate like he’s oblivious to us being in the beginnings of a fast take off. Welcome to the singularity…. See you in “10 years”…. 😉

Will be fun to look back on this in 2-3yrs. My bet: large aspects of programming will be abstracted away, but the demand's so high most devs won't mind. Most healthy devs anyway :)

grea video, love the IaC analogy pets/cattle

thank you!

what are you farming?

Likes

Such a great point about software going the same way as hardware infrastructure. The parallels are uncanny. A server used to be special and stateful. Now it's a completely expendable, disposable resource, just like a module of code written by an agent.

This short video was awesome, and I think we need more of it. Please keep posting videos about your thoughts. It's a great format. That doesn't mean I have to agree with you, but I really appreciate it.

I think infra teams are already shifting their focus to: 1. Internal integrations - skills, harnesses, MCP. 2. Enablement - getting non-engineers to ship product changes 3. Engineers shifting to harness development and quality control (e.g. fight the slop)

good one, please do more

Well said and right on the money with questioning the assumptions of cost as it relates to code production then and now. Traditional review of code at the speed of agents feels extremely antiquated. The need to have shared mental models doesn’t decrease though — that is the hardest nut to crack

you have a farm?

This guy gets it. For weeks now I have been refining what a software factory should look like. Today I kicked off an agent from my Tesla while waiting at the Dr's office.

Love this so much! (and your writings too ofc)

I also think dev tooling will evolve more towards agent-centric systems. Still we need to figure out good DX to interact with those then. Question is will GitHub adapt or not? I’m embracing this by building tools and infra for my agents and leaving little hole I can poke into.

Why it's looking like you switching to farming 🧺.

That's a really interesting take. Github's main role might change to just the collaboration with other people part. It will be as important as ever. Whoever will build that will take a huge chunk out of GitHub.

Just a heads up, if you are in Washington, then you aren't getting any vitamin D in the winter as the sun's angle is too shallow for the necessary UV radiation to penetrate the atmosphere.

Been thinking a lot about this since the start of the year. Code is just one bottleneck. We’ve left the ends untouched. Organizations were organized based on the way humans work. If you are going to bring these agents into an org you have to completely rethink the structure. And that is why startups who are agent first are such threats to these larger organizations that were setup pre-ai.

The whole industry is completely changing under our feet. Anything or anyone that wont adapt will be left behind.

The workflow layer is where the action lands

GitHub isn't dying; it's evolving from a passive warehouse to an active orchestration layer. The real question for '26 is where the 'source of truth' lives when agents are doing 80% of the commits. We're moving toward a paradigm where code is fluid, but architectural intent is the only permanent asset.