RE: @mgorny@social.treehouse.systems
I try not to think about this much and stay focused on my own work, but I can't deny that this has been a motivation killer for me.
Remote
Miguel Grinberg
@miguelgrinberg@mstdn.social
mastodon 4.7.2I write software. And from time to time, I write or speak about writing software.
816 Followers
99 Following
20 Posts
Joined December 01, 2022
Website:
Patreon:
Location:
Ireland
RE: @davidism@mas.to
Happy to see that at least one major Python project remains sane.
In this blog post, I share what I learned about the encryption algorithm that Bitwarden uses to keep secrets safe. https://blog.miguelgrinberg.com/post/how-bitwarden-encrypts-and-decrypts-secrets
Has anyone made a list of companies or organizations that forbid or at least do not encourage their employees to use coding agents or other generative AI tools?
New blog post: How to Host your Own Email Server https://blog.miguelgrinberg.com/post/how-to-host-your-own-email-server
New blog post: LLM Use in the Python Source Code https://blog.miguelgrinberg.com/post/llm-use-in-the-python-source-code
I have been testing other Linux distros because I expected this to happen, and now it is happening. Maybe I'll move to Gentoo, or perhaps Arch, still not sure, but what I'm sure of is that I'm saying goodbye to Ubuntu. https://www.phoronix.com/news/Ubuntu-AI-Features-2026
Here is how I create a React + Flask project in 2025: https://blog.miguelgrinberg.com/post/create-a-react-flask-project-in-2025
Replying to
@hugovk To be more clear, this policy is vague. There "Acceptable uses" section does not refer to submitting code at all. Under "Considerations for success" you can squint and maybe you see some indirect references that suggest it is allowed, disguised as general advice and not intended to be normative according to my interpretation.
Replying to
@hugovk But in any case, I have now updated the blog post to include a reference to this page. I should have included it, especially because I had seen it and did not fully answer the concerns that I'm raising. Thanks.
Replying to
@seanhelvey My policy wrt AI contributions is more blunt than Hynek's: https://github.com/miguelgrinberg/microdot?tab=contributing-ov-file#readme
Replying to
@danzin In that case they should make it clear in the policy that LLMs are used because there is an insufficient pool of contributors and that affects the project negatively.
They need to be transparent.
Replying to
@danzin Relying on how Claude Code signs commits for crediting is extremely sloppy, if I'm allowed to speak my mind. When researching material for my article I had to find those commits and even though it wasn't hard, it wasn't trivial either. And besides, you are implying that there are more usages of LLMs for larger changes, and presumably those do not have Claude as co-author. What if someone uses another LLM? Those need to be easily searchable as well.
Replying to
@danzin How would I know how significant is the part of the code that was generated with an LLM if there is no requirement to disclose every specific use? That is exactly the point I'm making in the article. I can guess that the use of coding agents is larger than those 8 commits, but because there is no disclosure requirement, I will never be able to calculate the percentage of code that was touched by these tools.
Again, if you are going to go down this road, you have to be transparent.
Replying to
@danzin My impression is that core developers are trying to have it both ways. I recognize that openly saying that the project accepts generative AI PRs is probably not what they want due to slop PRs. But I think it is unfair to hide that they are using it themselves. How spread is this? I'd really like to know now.
This is why I'm not moving my releases to Trusted Publishing and GitHub Actions. Say what you want about long-lived publishing tokens, but at least when using them I have full control of my release process and can take adequate protection measures.
https://nesbitt.io/2026/04/28/github-actions-is-the-weakest-link.html
The SMPTE catalog of media format specifications is now accessible for free at https://pub.smpte.org/doc/.
This is really big. If you are ever interested in parsing or generating video or audio files, you are likely to find official specifications here, the same ones the big companies use to build their media products.
Even if you are not interested in this yourself, I'm sure ffmpeg, VLC, and similar projects will greatly benefit from free access to this information!
Replying to
@sethmlarson@mastodon.social Great article. Somewhat related, I also find that the practice of setting version upper bounds in library dependencies is more often causing problems for users than helping them.