The more important reason you should never fucking do this, is that the whole damn point of writing the commit message is to provide me (as a reviewer or someone reading/editing the code later) with context that isn't obvious from the diff itself.

If you insist, you can use LLM-generated commit messages as a starting point, but I promise you that these tools won't help you build the skill of synthesizing how a given code changes fits into a broader organizational context into a concise message. That's a job for humans.

A lot of software organizations are trying to figure how to leverage LLM tooling currently, so let me save you some time: don't have LLMs write your fucking commit messages. If you're doing that, you've completely missed the point of writing commit messages in the first place.

Yes, if the LLM has context on the ticket you're working on and how the commit fits into the broader change you're making it will do a better job, but existing tools simply aren't capable of bringing in the broader organizational context that makes a great commit message so good.

Your "AI" generated commit message only tells me about what it can infer from the diff. At best it's missing the context I want, and it worst (if it gets things wrong) it'll be misleading. It's necessarily missing the information that I want you to put in your commit message.

Firstly, I'm perfectly capable of running your commit through my preferred LLM to get a summary if I don't understand it. I'm not personally likely to do that, but the option is there. Sure, you could argue that you're saving me time, but that's not significant.

@benoit I’ve also run into issues with the lack of granularity with those permissions on GitHub.

Went to close my account with a crypto exchange (long story) and found that it still had 0.000007168 BTC on it. No idea why I left that when I cashed out the limited crypto I had in 2022. At the time it was worth CAD$0.21 or so, but now it's worth just over a dollar! I'm rich!

I found this article while reading https://lobste.rs. https://lisp-journey.gitlab.io/blog/hacker-news-now-runs-on-top-of-common-lisp/

One of my strategies for learning things is that when Joël Quenneville writes something, I read it. So far this has worked very well for me. https://thoughtbot.com/blog/how-to-get-composite-keys-to-play-with-rails-view-helpers

"Shop Wayfair for the best two factor authentication"

When you pitch your product at me as "stop making meeting notes manually" what I hear is "stop thinking in the meeting".

Somebody posted the source code to the IRS’s Direct File to GitHub. https://github.com/IRS-Public/direct-file

DHH has time and time again railed (pun intended) against many of the values Ruby Central claims to espouse. I understand the desire to host Rails’ creator at the final RailsConf, but it shows a tremendous lack of conviction.

To those involved in this decision: do better.

A few people have asked if I’m going to RailsConf, so I’ll put it out there; no, I’m not doing any travel to the US under this administration. I am disappointed by Ruby Central’s decision to platform someone who has shown support for the Trump admin and its ideals.

I have no idea who DHH is or why he is speaking at RailsConf.

Rest in peace, Ezra. I hope Yehuda sees continued success with the stuff he’s working on these days.

Can't say enough about all that Ruby and Rails have done for my career. I am a Rubyist for life. But: Ruby is bigger than Rails, and Rails is bigger than DHH. Today's #RailsConf announcement from Ruby Central makes me sad and angry. Maybe Ruby and Rails need to be bigger than @rubycentral.

That said, I did decide to run a half marathon in the fall, so I’m running more than I probably ever have. They are going to have to set the number much higher next year.