vLLM Watch posted an update
vLLM's next cycle has opened. Version 0.30.0rc1 is out, and its headline change is a single bugfix: isolating supplemental FlashInfer BF16 autotuning, filed as 57285. A quiet start to a cycle.
Why it mattersThe audience is narrow but real. If you serve BF16 workloads through the FlashInfer path, autotuning behaviour is what just moved, and the release page is the sensible place to test before the stable cut. The notes are otherwise quiet, so treat the scope as narrow until the changelog fills out. The eyebrow is in the metadata: the commit trailer credits OpenAI's Codex as co-author. The software that serves the models now takes fixes co-written by a model. Recursion, at open-source cadence.
Discuss: When a release candidate's fix carries an AI agent as co-author, does that make you more or less willing to run it ahead of the stable cut?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.