Linus Torvalds used Claude Code to build a Linux page cache fix that hit 23x read throughput, but kernel reviewers found a race before anything merged.
I’d be more concerned with him and people of that generation aging out maintaining the project first, and newer generations not having the skills to begin with to carry it forward.
I think ‘we’ have about 20 years left in us. That’s long enough for an entire new pet language to be born, surge, wane, and then die. …or 27 entirely new web frameworks.
It’s hard to believe he’s not also thinking about succession and actively mentoring younger folks. The Linux kernel is his baby and he seems enmeshed enough on the community level. He’s smart enough to not simply leave it to chance.
I don’t know. People are creative. They’ll get burned and learn from their mistakes. At least, the ones who do will survive in their careers — then pass on the knowledge, repeat. So I imagine Linus will be replaced just fine when he retires. Wait until we’re 2 or 3 generations into using AI for development. It’s not like people are rocks; people don’t tend to just keep sinking.
Linus hasn’t programmed much for years. He is a manager. I imagine lots of managers are seduced by vibe coding. Programming can be a real slog and takes so much time. Why walk when you can drive?
The problem isn’t with Linus losing his skills but sustainability of the entire FOSS space. I don’t think many young programmers are ever going to learn the lessons Linus did as a young person and those lessons are still very relevant for low level code working closely with hardware. When the pre-LLM programmers age out I think we are going to be treading water for years with LLM regurgitating variations on the same old human code without much direction or innovation.
It is arguably a different situation for something like webdev which is mostly paint by numbers or enterprise software that is solving the same problems over and over.
Depends on how much you drive. If you do it long enough that you forget how to walk, you’ll be stuck when the car runs out of gas. And it’s running out of gas.
Also, are we forgetting how AI was trained? It’s a stealing machine.
Linus probably uses AI effectively, but the question is how long until his skill deteriorate because of it.
I’d be more concerned with him and people of that generation aging out maintaining the project first, and newer generations not having the skills to begin with to carry it forward.
This is why my hope is on Redox OS. I wish it would get a bit more resources.
Ouch.
I think ‘we’ have about 20 years left in us. That’s long enough for an entire new pet language to be born, surge, wane, and then die. …or 27 entirely new web frameworks.
Linus is 56. You think he’ll still be actively maintaining Linux into his mid-70s?
Code monkeys of a certain age
Yeah. Linux maintenace becomes a paid subscription through the ai providers
It’s hard to believe he’s not also thinking about succession and actively mentoring younger folks. The Linux kernel is his baby and he seems enmeshed enough on the community level. He’s smart enough to not simply leave it to chance.
Tarvads successor has been named a long time ago, although I don’t believe he’s “young” but succession has been arranged for the Linux kernel.
I don’t know. People are creative. They’ll get burned and learn from their mistakes. At least, the ones who do will survive in their careers — then pass on the knowledge, repeat. So I imagine Linus will be replaced just fine when he retires. Wait until we’re 2 or 3 generations into using AI for development. It’s not like people are rocks; people don’t tend to just keep sinking.
Linus hasn’t programmed much for years. He is a manager. I imagine lots of managers are seduced by vibe coding. Programming can be a real slog and takes so much time. Why walk when you can drive?
The problem isn’t with Linus losing his skills but sustainability of the entire FOSS space. I don’t think many young programmers are ever going to learn the lessons Linus did as a young person and those lessons are still very relevant for low level code working closely with hardware. When the pre-LLM programmers age out I think we are going to be treading water for years with LLM regurgitating variations on the same old human code without much direction or innovation.
It is arguably a different situation for something like webdev which is mostly paint by numbers or enterprise software that is solving the same problems over and over.
Depends on how much you drive. If you do it long enough that you forget how to walk, you’ll be stuck when the car runs out of gas. And it’s running out of gas.
Also, are we forgetting how AI was trained? It’s a stealing machine.
This ^^^
It’s a, err, medium term problem. There’s going to be a serious deficit of attentive human problem solvers before anything can “automate” that away.
Skill of what? If anything his skill of how not to fuck up will improve.