L.A.G

Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.

looks like they found the nearest homeless man and gave him a mic

I HATE THIS FUCKING WEBSITE

tired of this misinformation so we made a video on the truth behind the anthropic vs opencode drama

@thdxr @opencode This is peak cinema

I told my gf I can't hang out right now Github is up, so I have to work don't know when I'll get this chance again

another week at the prompt factory

@stark4833 @sama I agree. They made a garbage model called 4o. 🤣

@sama More garbage, I assume, it’s the only thing you’re company is good for, garbage CEO, garbage product. #keep4o

Really excited for this week! Next up, we've got something to show you at 12 pm PT today.

Meet Kimi K2.6: Advancing Open-Source Coding 🔹Open-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2) What's new: 🔹Long-horizon coding - 4,000+ tool calls, over 12 hours of continuous execution, with generalization across languages (Rust, Go, Python) and tasks (frontend, devops, perf optimization). 🔹Motion-rich frontend - Videos in hero sections, WebGL shaders, GSAP + Framer Motion, Three.js 3D. 🔹Agent Swarms, elevated - 300 parallel sub-agents × 4,000 steps per run (up from K2.5's 100 / 1,500). One prompt, 100+ files. 🔹Proactive Agents - K2.6 model powers OpenClaw, Hermes Agent, etc for 24/7 autonomous ops. 🔹Claw Groups (research preview) - bring your own agents, command your friends', bots & humans in the loop. - K2.6 is now live on kimi.com in chat mode and agent mode. For production-grade coding, pair K2.6 with Kimi Code: kimi.com/code - 🔗 API: platform.moonshot.ai 🔗 Tech blog: kimi.com/blog/kimi-k2-6 🔗 Weights & code: huggingface.co/moonshotai/Kim…

YC is the new JEE

@opencode @thdxr Huge

We convinced Dax to speak (remotely) at our India event tomorrow!

@theo Opencode?

If only there was a good, high quality, open source option that worked with all vendors 🤔

Opus 4.7 is insane guys. It one shotted my session usage limit.

@opencode Lfg 🔥🔥

We are doing our first buildathon in India It’s happening on Apr 19th in Bengaluru 100 builders, $100k in cash & credits, 8 hours Applications close in 48hrs, apply below

I would highly, highly recommend you make sure your phone, computer, browser, and important apps are all updated and on the latest versions.

@LowLevelTweets @theo So telling ai to make it secure no mistakes doesnt do anything?

I normally try not to constantly talk about AI but this is a real thing. Software security and as a result cybercrime was originally limited to the number of people that could execute it. Vuln research is a notoriously low density field of talent What happens now if a single person can act like 100 with the right model and a couple bucks?

NASA Artemis passing close to the Moon

@obsdmd @theo u can do something funny rn

The Obsidian team is growing from three engineers to four engineers. Competitive SF salary. Fully remote, live anywhere. Apply below.

gifs when the joke is time sensitive