Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional…
ideal prompting technique is: - thin prompts - thick artifacts + context - thin skills
How we built our knowledge base Authors: [@hi_im_isaac_](https://x.com/@hi_im_isaac_), [@learnwdaniel](https://x.com/@learnwdaniel), [@gaozenghao](https://x.com/@gaozenghao) note: the interactive version of full technical blog available: https://www.cerebras.ai/blog/how-we-buil…
It's open-source now! Have your codex set it up for you and import your clothes. link below https://t.co/1xtpdHeBGM
codex is really good at writing code on tldraw https://t.co/vJUuMmhmgw
Kimi K3 is really, really good. https://t.co/CCjSJvBwWK
The Self-Driving Company We are beginning to see what happens when a company learns to operate itself. In the past six months, engineers at Replit have nearly tripled code output. Review times held steady. Reversions and product incidents have stayed flat. Quality metrics impr…
https://t.co/uFkkNhc2Dx
We're training neural nets in tldraw offline
https://t.co/enLPTRYfFB
made a tiny terminal app to download videos from x, youtube, instagram, etc copy a post's url, paste it, yoink it npm i -g yoinks https://t.co/qLEzvQtCsq https://t.co/oZDJyePqQ1
How do you make 768 servers look like 1? The answer is sharding. I wrote about *why* it's needed and *how* it's achieved for Postgres and MySQL. It's a fun one. Link below. https://t.co/Bx2W0GGEVZ
I got "how did you make these diagrams?" a lot yesterday. BTS time! Most began their life as static drawings in Excalidraw. It's helpful to nail down layout and visual goals before jumping into animation. Here's what they looked like: Next step, Cursor 🧵 https://t.co/xFwaXS…
https://t.co/CtlgZTvQAJ
Benchmarks are dead (for us). Our RSI system makes benchmarks too easy. Give ours a benchmark and it builds its own solution–then beats SOTA. We just did it on 6 at once: math, coding, planning, long-context, tool use, web apps. No human tuning. https://t.co/Jbh0A0MAbg
I'm working on some new sections for the Solo homepage and needed new illustrations. Instead of having Sol try to one-shot it, I had it build me a little studio. It can still access the raw SVG, but I also have knobs I can tweak. Now we work iteratively until it looks perfect!…
https://t.co/lojJhq1cxA
hear all about tldraw offline https://t.co/ecMtPnec7H
New Max Agency with @FactoryAI CTO & Co-Founder @EnoReyes. This was a particularly fun one. Since we're both building harnesses, we nerded out pretty deep. Awesome conversation on why the harness matters more than the model running underneath it, how Factory built Missions, + …
codex-tldraw-mcp now keeps repo diagrams useful as code changes! it creates trackable JS/TS module graphs, detects changed, stale, and new elements, and marks drift on .tldr boards. GitHub: https://t.co/jPmbnIoSAW npm: https://t.co/ZS4VYAZxH7 would love to hear feedback!
Ranked by when the bookmark was saved during this London date range. Interactions are likes, reposts, replies, and quotes combined. Metrics are X's latest public counts; older items can be stale once they leave the recent-bookmarks sync window.