My first vLLM contribution just shipped in v0.22.0 🎉
GatedDeltaNet Marlin TP≥2 fix (#36329)
Quantized Qwen3.5 at TP≥2 was crashing on Marlin's 64-element minimum tile width. Fix: replicate the one tiny projection layer that produces 64-wide vectors.
@K_Stanowski@K_Stanowski - serio..., w dobie gdzie jest ksef, ai, pełno narzędzi i procesów do "monitorowania kowalskiego" mamy problem kilometrówek... posłowie mają za dużo przywilejów.
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
Yes, it’s mostly nature — you usually have to BYOC since there aren't many women to meet there! 😆
winter season - ⛷️
summer - bike, hiking, campfire, wilderness
You mentioned weekends - family etc for me Dolys are the place were focus shifts to the one next to you or the nature.
Mieszkanie w mieście powiatowym niedaleko Warszawy to najlepsze co może być.
Jedziesz sobie do centrum miejscowości bo powiedzmy chcesz zjeść chińczyka. W Warszawie byłaby to mordęga a w Żyrardowie:
- podjeżdżasz pod sam lokal, miejsc w opór wszędzie, nie trzeba iść pół kilometra z najbliższego
- zero SPPN
- zero SCT
- miejsca normalnie skośne a nie równoległe jak wszędzie w Warszawie
- zjadasz chińczyka
- wracasz do domu co zajmuje 5 minut a nie godzinę
Piękna sprawa
PS. No i Twoje podatki idą na potrzebne rzeczy a nie na zwężenia w Warszawie
Over the past month, some of you reported Claude Code's quality had slipped. We investigated, and published a post-mortem on the three issues we found.
All are fixed in v2.1.116+ and we’ve reset usage limits for all subscribers.
New in Claude Code: /ultrareview (research preview) runs a fleet of bug-hunting agents in the cloud.
Findings land in the CLI or Desktop automatically. Run it before merging critical changes—auth, data migrations, etc.
Pro and Max users get 3 free reviews through 5/5.
@adam_jesion@bcherny@llmdevguy In last 2 weeks I have shifted with lots of my workflow in to the app. Which I find is now more reliable/predictable than CC
Yes, that could be a trigger. I have noticed, that more they rollout tools like Routines, more "old fashion" ways are risky. I'd use CLAUDE.md: to define strict rules for what the agent is allowed to do.
Isolate the environment: Run agents/loops inside Docker containers.
Limit tools: Explicitly list allowed tools (e.g., read-only) and avoid --dangerously-skip-permissions.