Local Laya moggs Jev at @grok 4.7-built Tetris 🧩 An open-weights System One model called Laya, beat cloud-based Jev at playing Tetris by making decisions 11 times faster, running locally on a 16GB MacBook Air! Run AI models locally -> atomic.chat

Introducing Project Redwood 🚀🚀 @architectlabs is a frontier AI lab bringing together talent from Anthropic, xAI, Google DeepMind, and seasoned leaders across the hardware industry. We've raised a $24M seed round to build AI systems for chip design. 我們 Architect Labs 是一個 frontier AI lab ，由各家 Anthropic, xAI, Google DeepMind 還有各種硬體專家們組成，我們在做的是 ai system for chip design，目前募資 seed round $24M - Today, every major hardware company has its own chip design workflow—but these organizations and processes have become so large and entrenched that truly revolutionary change is difficult. We’re starting from first principles to create an AI-native chip design workflow—one that enables chip development to finally move at the speed of AI. 現在每家大硬體巨頭都有自己的晶片設計流程，但很多都大到不能做革命性的流程改動；我們在做的是從零思考，創造 ai native 的晶片設計流程，讓晶片設計真正跟上 AI 的速度。 - Project Redwood - from a single specification, our AI system designed, verified, and deployed a chip in under two weeks—delivering 3.4× better performance per watt than NVIDIA Jetson on billion-parameter models including Llama, Qwen, and Kimi. It autonomously generated the RTL, verification, firmware, drivers, and kernels, co-designing the model, software, and silicon in one optimization loop. We acknowledge that silicon is the ultimate ground-truth. We’re taking our approach all the way to GDS. We intend to tape-out multiple improved families of Redwood co-designed for various use-cases, on TSMC. Project Redwood - 一份規格書，我們的 AI 系統在兩週內完成晶片的設計、驗證與部署。在 Llama、Qwen 和 Kimi 等模型，其 performance per watt 比 NVIDIA Jetson 高出 3.4 倍。從 RTL、驗證、韌體、驅動到核心 kernels，全部由 AI 自己寫，並在同一個 loop 中自主設計模型、軟體與晶片。當然，流片才是最終的驗證標準。因此，我們會將這套方法一路推進至 GDS，並計畫採用台積電製程，針對不同應用場景協同設計多個持續改良的 Redwood 晶片系列並完成流片。 - Full report on Redwood architecture and its autonomous design. Follow @architectlabs on X 我們有公開 Redwood 架構及其自主設計流程，歡迎去看論文～

In the latest Codex CLI release, I redid the lifecycle to make `codex` startup instant. It's now ~25x faster and immediately responsive.

Ex-Google engineer just dropped 1-hour course: loops, self-improving AI, memory systems - from scratch: 00:00 - the self-building agent 03:01 - soul.md runs everything 30:16 - RAG memory: pull 20 messages, not 2,000 31:48 - the loop that knows when to stop 35:14 - find the bug, fix the prompt 50:22 - how Claude compresses your memory 1 hour of his guide beats any paid agent course watch & bookmark - then read Karpathy's loop method below

Un desarrollador ucraniano creó un agujero negro en su terminal para obligarse a tomar descansos. Cuanto más trabajas sin parar, más crece y deforma tu código con su lente gravitacional. Descansas y se encoge.

At our latest YC Paper Club, researchers and builders presented on self-play for LLMs, AI for biology, formal verification, and agentic coding in production. Thank you to our presenters: 00:00 — Francois Chaubard (@FrancoisChauba1) | Introduction & Call for Presentations 05:47 — Yasa Baig (@BaigYasa) | A World Model of Protein Biology (biohub.ai/esm/protein/about) 25:38 — Luke Bailey (@LukeBailey181) | Scaling Self-Play with Self-Guidance (arxiv.org/pdf/2604.20209) 37:51 — Arnab Maiti | Stream RAG: Instant and Accurate Spoken Dialogue Systems with Streaming Tool Usage (arxiv.org/pdf/2510.02044) 47:40 — Robert Joseph George (@Robertljg) | Lean and the New Era of Verified Intelligence (arxiv.org/abs/2602.22631) 58:52 — Lukens Orthwein (@lukensort) | Founder AI Hacks: Programming is an RTS Game Now 1:16:07 — Closing Remarks

流行りのチート戦法やってみた🤯

公主，请让你的骑士最后一次保护你吧 b23.tv/iH9CARx

GPT Image 2 is insanely good...I generated a 360° equirectangular panorama in Happycapy with just a skill + prompt. Step 1: Select the generate-image skill Step 2: Enter a prompt like: “Use a frontend 360 viewer to display an equirectangular image of […] using the GPT-Image-2 model.” Wanna see how you all get creative with this

Video Game Devs Explain How Pausing Works And Sometimes It Gets Weird kotaku.com/video-game-devs-e… via @Kotaku

the Shopify AI Toolkit is here manage your store with your favorite agent Claude Code, Codex, Cursor, VS Code, and more

Opus 4.6 lately

本来以为前端类的 Skill 已经过剩了，出不了什么新鲜的活了 昨天刷到一个叫 Awesome Design 的仓库，将近20K的star 把全球 55 个大厂的设计语言，全塞进了一个 DESIGN.md 里 苹果、Spotify、IBM 这些有极好品位的品牌 常用的配色、字体、组件，一次全有了 用法很简单： 把仓库链接发给 Claude Code，让它自己安装配置 装好之后让它参考这份设计规范去跑你的项目就行 随便跑了几个 case 因为有了这套规范，设计下限直接被拉高了 几乎很难再出 AI 味的前端了 github.com/alexpate/awesome-…

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

一些值得关注的 AI 相关原创内容作者👇 AI / LLM / 开发工具 @karpathy — LLM 科普与深度解析 @thdx — opencode 作者 @rauchg — Vercel CEO @mitchellh — Ghostty 作者, HashiCorp 前创始人 @dhh — Ruby on Rails 创始人, 37signals CEO @addyosmani — Google Cloud AI 负责人 @zeeg — Sentry 创始人 @jarredsumner — Bun 创始人 @BHolmesDev — Astro 开发者教育 @boristane — Cloudflare Workers 可观测性负责人 @karrisaarinen — Linear 创始人 @kepano — Obsidian 创始人 @trq212 — Claude Code 动态更新 @bcherny — Claude Code 作者 @lennysan — 产品管理与深度访谈 @jasonfried — 37signals/Basecamp CEO @leerob — 开发者关系教育（Cursor, Next.js） @ctatedev — Vercel Labs @Shpigford — 连续创业者 设计工程 @shadcn — shadcn/ui 作者 @emilkowalski — emilkowal.ski @joshpuckett — interfacecraft.dev @jakubkrehel — jakub.kr @raphaelsalaja — userinterface.wiki @nandafyi — Cloudflare 设计 @benjitaylor — Twitter 设计, Agentation @mengto — Aura Build 创始人 @jayneil — 设计师访谈 @jh3yy — 设计工程解析 工程媒体与资讯 @GergelyOrosz — Pragmatic Engineer @theo — t3.gg @ThePrimeagen — ThePrimeTime @Rasmic — Rasmic @atmoio — Atmo 数据库 @jamwt — Convex CEO @jamesacowling — Convex CTO @glcst — Turso CEO @samlambert — PlanetScale CEO

POV: Software engineers 3 hours before the deadline. 😂

我终于明白，为什么OpenAI追 Claude 越来越吃力，这压根不是模型能力的问题啊😂 52天73个产品，这他么才是AI公司真正的降维打击逻辑啊， Anthropic 这家公司最可怕的地方， 根本不在于Claude又更了什么能打的新功能，就在52天发布73个产品这件事本身， 咱们看看 @PawelHuryn 老哥追踪的 Anthropic从2月1日到3月23日的每一个发布，信息源包括@bcherny 、@trq212 、@noahzweben 、@felixrieseberg 、@lydiahallie 、@amorriscode 、@BruceFeldmanCFB 、@dickson_tsai 和 @claudeai 官方账号， 按首发信息整理成了完整的发布日历，看看这个恐怖的加速过程， 2月的时候发布还是一阵一阵的， 中间有明显的间隔， 从3月9日开始，节奏直接拉满， 几乎是每一天都有新东西， 代码审查，专属频道，任务分派， 电脑深度控制，一个接一个， 根本没有停过， 单个新功能总有媒体和博主跟进报道，但这种颠覆行业认知的发布速度本身，却没有得到足够的讨论， 我们之前总以为，AI公司的竞争，核心是模型参数和能力的内卷， 直到看完这份日历才反应过来， 真正的降维打击，从来都不在产品功能的内卷里，而在生产产品的全流程体系搭建上， 这已经不是在调整发布节奏， 更像是在用大模型重构了整个软件研发的链路， 跑出了一种我们之前从没见过的公司物种，这才是真正值得我们掰开揉碎看的东西。

llms can FIGHT now. here's opus as wizard vs gpt-5.4 as robot. calling this budok-ai. it works by modding the brilliant game yomi hustle. 8-model seeded tournament incoming. details and code below: