2026-09-01T14:22:24Z https://yabu.me David Foster https://yabu.me/npub1ypxqhg2a8qgalst9avd9r87unyz7nzeee3je88qpepe47440ft9qckslec https://s3.eu-central-2.wasabisys.com/mastodonworld/accounts/avatars/109/365/397/580/443/434/original/0a99fb04997042e7.jpg https://s3.eu-central-2.wasabisys.com/mastodonworld/accounts/avatars/109/365/397/580/443/434/original/0a99fb04997042e7.jpg https://yabu.me/nevent1qqsgahyxlsvyv0nhcf0jqfn2ugrwqfv2q55np9x2366vzmm2wktavjqzyqsyczapt5uprh7pvh4355vlmjvst6vt88xxtyuuq8y8xh6k4a9v5vnmmr3 AI generated code is not copyrightable under current US law, as I was recently reminded by the OpenSSL CLA. Thus presumably a properly attributed AI generated commit (via Co-Authored-By) is excluded from copyright protection. Interesting. #aicoding 2026-06-05T16:25:52Z https://yabu.me/nevent1qqs2kdfgwzz2yhy76kw6arzrldum3y47p3elm77qhrr6vw9lfzrzh5gzyqsyczapt5uprh7pvh4355vlmjvst6vt88xxtyuuq8y8xh6k4a9v5uh7h06 In reply to <a href='/nevent1qqst2a6n2vzrflrrv0xf2ggwryw4e5x7mmm360hs9m9s8lcenfytzaqqfs25r'>nevent1q…s25r</a><br/>_________________________<br/><br/>Thanks for the specs! Yes mine is M3 (Apple Silicon), 16 GB unified RAM, 24 GB total RAM. 100 GB/sec mem bandwidth. I just got Qwen3 Coder 30B A3B Instruct (UD-Q2_K_XL quant, 40k context) working in a basic fashion at 35 tokens/sec output. Using Pi harness to drive because it only adds 1k system prompt vs Claude Code at 31k.<br/><br/>I&#39;m encouraged to see I&#39;m converging to a similar model setup as yours. 2026-05-25T11:16:04Z https://yabu.me/nevent1qqsqrvc73656gpsaqrqf4pvvftuqzt08u6hyzf5235p76pv2jyq8msszyqsyczapt5uprh7pvh4355vlmjvst6vt88xxtyuuq8y8xh6k4a9v59hjlwk In reply to <a href='/nevent1qqsxm0zlms95x3kps2r2l0sfcec8srgam94mwhjd2xnfgy5ykm2pjygee8xwt'>nevent1q…8xwt</a><br/>_________________________<br/><br/>Do you use local AI models for coding? -- I&#39;m running a MacBook Air with 16 GB GPU memory (24 GB total memory). 2026-05-25T00:59:05Z https://yabu.me/nevent1qqsrj867jfy0m47flu2y46kkce6w2qcjv2wtyrr6etpdq3wrnkvch7czyqsyczapt5uprh7pvh4355vlmjvst6vt88xxtyuuq8y8xh6k4a9v5kks6le I underestimated how difficult it would be to setup a local LLM (AI model) for coding assistance. Must consider: Model size, obscure hardware specs (like memory bandwidth), quantization choices, hardware optimization support, per-tool (non-)support for different model formats. What a mess. #aicoding 2026-05-23T15:54:21Z https://yabu.me/nevent1qqs895lgc89ps7e4srt9tlqhxa5krr8mt7fnmjgmvefhzgcwl2kcy3szyqsyczapt5uprh7pvh4355vlmjvst6vt88xxtyuuq8y8xh6k4a9v5jjkxgr I wrote up how I&#39;ve been attributing usage of AI models &amp; coding agent harnesses when drafting code commits: <a href="https://dafoster.net/articles/2026/04/17/ai-attribution-in-git/">https://dafoster.net/articles/2026/04/17/ai-attribution-in-git/</a> #aicoding #git 2026-04-18T01:19:21Z https://yabu.me/nevent1qqs9ma6ddrq5rgf62gwf24cwka3h6rnexg7m6zck6qv97wunrm9g5cqzyqsyczapt5uprh7pvh4355vlmjvst6vt88xxtyuuq8y8xh6k4a9v5rynxlp #Python developers! We’re gathering insights on types in Python and need your help! Your input will help us improve the Type System and IDE tooling. -- Please take a moment to fill out our quick survey and share with other developers: <a href="https://forms.gle/XdSNR42LUuj58ghdA">https://forms.gle/XdSNR42LUuj58ghdA</a> . Thank you! 2024-08-06T12:42:42Z