Lin

@ChujieZheng 27b please!

We are planning to open-source the Qwen3.6 models (particularly medium-sized versions) to facilitate local deployment and customization for developers. Please vote for the model size you are **most** anticipating—the community’s voice is vital to us!

@OpenAI I really want to try pro plan

Today, we closed our latest funding round with $122 billion in committed capital at an $852B post-money valuation. The fastest way to expand AI’s benefits is to put useful intelligence in people’s hands early and let access compound globally. This funding gives us resources to lead at scale. openai.com/index/accelera…

@thsottiaux Thank you

Our Codex dashboards are showing increased rate of users hitting rate limits and since we don't fully understand why I have made the cautious decision of resetting the usage limits for all plans. Enjoy. I also wanted to celebrate us finding a pocket of fraudulent accounts that we banned and have helped us regain some compute. The fight against abuse never stops, but it's important to mark the moment and make it a little shared victory.

@uops_info Finally

I have added latency, throughput, and port usage data for Emerald Rapids, Meteor Lake, Arrow Lake, and Zen 5 to uops.info/table.html.

@awnihannun Congratulations Awni! you are THE GOAT!

I joined Anthropic as a member of the technical staff. Excited to work on frontier modeling at a place with unwavering values and a generational mission.

@mikehostetler I'm so excited! Show them BEAM's WAY(tm). There's already an awesome book about this: link.springer.com/book/10.1007/9…

Jido will have a true self-evolving agent soon Not just extra SKILLS.md - that's childs play Soooooooon

@chris_karani Great insight. One interesting similarity is that AGENTS.md is much more like traditional UNIX Makefiles. These handle the behaviour of the entire subtree and thankfully this agent restores us to the UNIX ethos of "everything is a file".

@ashvardanian Your naming is spot on.

Remember StringWars? It's time for NumWars now, and the stats for i8 → i32 & bf16 → f32 matrix multiplications are amazing! MKL-level performance in just a 5 MB binary, 6x smaller than NumPy with OpenBLAS... + for C, C++ 23, Rust, Go, JS, & Swift github.com/ashvardanian/N…

@chris_karani You are the GOAT!

spent today finally getting Qwen GGUF correctness over the line in Espresso on Apple silicon. the headline result: fresh cold-start still works. late prefix `[9707, 21806, 11, 358, 2776, 14589, 369, 279]` now gives `3681`. full `Hello` continuation now matches raw GGUF too. 🧵

@rachpradhan This is incredibly cool.

I replaced FastAPI's entire HTTP core with Zig. Same decorator API. Same Pydantic models. 7× faster. 47,832 req/s vs FastAPI's 6,800. 2.09ms p50 latency. Introducing. TurboAPI. Here's the story..

@LottoLabs I can't wait! You've really enlightened me about this 27b gem.

Qwen 27b remains my favorite still, gonna do a write up on all the models tested so far

@gonizahavy @exolabs @HotAisle @Prince_Canuma @PixelTrik07 @alexocheema That's amazing! The performance can be improved later but I'm so super-impressed it works!

@lin72h @exolabs @HotAisle @Prince_Canuma @PixelTrik07 @alexocheema Currently it’s still useless with qwen3 0.6b running at 7 tps 😆 But there is progress, yesterday it ran at 0.9 tps !

MLX on ROCm?? 👀 EXO on Strix Halo?? 👀 It’s nowhere near ready, but it works! @exolabs @AIatAMD @HotAisle @zcbenz @shshnkp @NiketanNripesh github.com/ml-explore/mlx… + github.com/NripeshN/mlx/p…

@gonizahavy @exolabs @HotAisle @Prince_Canuma @PixelTrik07 @alexocheema that's amazing, does it mean i can use the mlx finetune on my vulkan based gpu?

Apparently I’ve lost it and decided to implement Vulkan backend instead 😱😅 Still early, but it works. If someone is interested in sponsoring, it will definitely accelerate development ⛏️ github.com/ml-explore/mlx… @exolabs @HotAisle @Prince_Canuma @PixelTrik07 @alexocheema @nummanali @ibnAmjid @zcbenz

@dkundel thank you

Thank you for your patience everyone 😊 limits are reset!

@ashvardanian This is massive!

Can’t believe it — all CI is finally green. About to merge the biggest open-source contribution I’ve ever made. The PR was opened in 2024… Meet NumKong — StringZilla’s big brother. Same monstrous DNA 😅 but built for tensors rather than strings! Link ⬇️

@anteasuncion Keep up the awesome work!

tackling swift driver temp directory issues on freebsd

@anteasuncion Thanks for working on freebsd. hope there will be a 15.0 version

building swift on freebsd 14.3

@__morse Nice

this is a TUI. just added a new 𝙷𝚎𝚊𝚝𝚖𝚊𝚙 component to termcast.app

@pontaro_pg You've tackled top2 feature requests from FreeBSD community

OpenClaw on jail #FreeBSD 15.0-RELEASE Install notes (Draft) github.com/isaponsoft/fre…

@pontaro_pg thanks for sharing, what an achievement !

@lin72h I wrote installation notes on GitHub, so I hope they will be helpful. github.com/isaponsoft/fre…

Success! 🚀 llama-server w/ CUDA on #FreeBSD 15.0 Linuxulator! Stack: RTX 4060 Ti 16GB Rocky Linux 9 chroot dummy-uvm.so hack Perf: Qwen2.5-7B at 227 tok/s! ⚡️ Build notes & procedure will be on GitHub soon. Stay tuned! 😈🔥 #LLM #CUDA #NVIDIA #FreeBSD