This is exactly right. Too many people directly copying SpaceX’s homework and saying everyone should vertically integrate at all costs. But in Connor’s words, “rebuilding the industrial base inside every new prime makes little sense.”

I am cautiously optimistic that we will be able to run the @SpaceX version of the VR72 at close to 250kW average power, meaning peak power would be ~10% higher. This is a big deal if it pans out and would apply to our ground-based systems too. The close working partnership and co-design with the amazing @Nvidia engineering team is key to achieving this.

thanks for sharing!

Russia plans to conduct a launch abort test of its next-generation crew spacecraft next year? I'll believe that when I see it. arstechnica.com/space/2026/0…

Russia is rushing to establish a copycat of Starlink’s satellite constellation and is speeding up ballistic-missile production. By one estimate it can now launch over 100 of all types every month

The House would be in session now, but GOP leaders sent us on recess two weeks early to avoid a vote on releasing more Epstein Files. Congress won’t reconvene until after the election, at which point @RepTeresaLF @RepRoKhanna and I will force that vote.

Personal AI is in its Homebrew computer club era but it won’t stay there Truly personal AI will mean you need to own your own skills and memory

Boeing’s price for keeping Starliner flying? $359 million. arstechnica.com/space/2026/0…

Peter Steinberger, the creator of OpenClaw, open-sourced his entire agent setup The core idea is one folder that every agent on his machines reads from. A single AGENTS.MD holds his rules, and a script symlinks it into ~/.claude/CLAUDE.md and ~/.codex/AGENTS.md This way Claude Code and Codex always follow the same instructions. Every other repo gets one line at the top: READ ~/Projects/agent-scripts/AGENTS.MD BEFORE ANYTHING. Change a rule once and every project picks it up. Skills work the same way. 69 of them live in one place, each with a short description the agent reads to decide what to load, and scripts/sync-skills links them into both agents. Fork it, replace his rules with yours, and delete the skills you do not need. His AGENTS.MD is full of his own hosts and accounts. 6.6k stars, MIT - github.com/steipete/agent-sc…

Florida Attorney General James Uthmeier has asked a court for an emergency injunction against OpenAI that would block all new model development without independent third-party oversight, and he quotes Paul Christiano's statement on joining tne OpenAI board in the filing.

红杉的一位合伙人 Pat Grady 昨天直接把他们和自己的一个 LP meeting（Boston College）的内容录制发了出来，信息密度非常高，强烈推荐大家去看一下： - JEV 在过去的 7 天里从零增长到了 100 million 的收入。在此之前 The Information 刚刚爆出来，他们现在估值是在 10 billion。 - 现在 AI 公司的增速，比历史上的都要快很多。Instinct 发布以来，保持了 day over day 10% 的增长率。 - 200 到 700，指的是某一个 high-value knowledge work company 从去年年底 200M 增长到今年预测 700M 的收入。（后面那个 2 到 50 和 0 到 70%，他说还是跳过吧，可能还是有些敏感。） - 在 AI 公司，现在有一个连融两轮的行业惯例或风气：把"陪你建公司的"和"只出钱的"拆开。他拿红杉内部的数据（过去 12 个月里的 7 个案例）来举例： 红杉作为陪建的那一方进场，平均的投后估值是 1.1 亿；而仅仅一个月后，只出钱的那一方给的下一轮平均估值就达到了 34 亿。他自己的说法是以前从没见过，这就是泡沫。 - 很多企业想 own 自己的 intelligence：基础模型厂商只给你 Pareto frontier 上有限的几个点（大中小几档），但企业自己的 workload 需要的往往不是这几个点，所以要自己训、自己部署、专门为自己的 workload 优化。这也是为什么他说专用架构的 lab 在跑赢通用的 lab。 - 应用层公司每 4 个月就要 reinvent 一次自己，因为底层模型一直在变，地板一直在抬。 - 组织从 hierarchy、command and control 往 network of agents 走：AI 负责信息流转，人相对自主地工作。他说没有 Jack Dorsey 几个月前那篇帖子说得那么夸张，但方向就是这个。 - labs 内部：模型去年大部分时间已经在自己造自己，最近才意识到 alignment 没被当回事；都在押 custom silicon、新架构和 continual learning，算力全员紧缺；为了抢 API token，一边降价，一边派 consultant 去给 Fortune 500 定制吃 token 的工具 - 模型能力和实际落地之间有很大的 diffusion gap，这就是应用层的机会。 - hyperscaler 今年开始借钱做 capex，不再靠 free cash flow。 而且我特别喜欢他这种说话方式，他每句话就是只说一遍，说过去就过去，所以信息密度很高。推荐大家可以直接去看原视频， 总共视频时长只有 15 分钟：loom.com/share/c016702964a04…

A justification for "exclusion zones" on the Moon.

Xerox, if they still existed, would be even more upset by what I've written recently about another PARC project. It will be part of the next volume of my MAINTENANCE book. It is online now, here: books.worksinprogress.co/boo…

Apple is lost because bean counters and logistics men can't navigate paradigm shifts. Happy they put a hardware man back in the hot seat, but it's the software that needs fixing.

I don’t bet on this garbage because I think it’s terrible for society, and in the case of space it would be unethical. But I have to say that a launch of a Delta ship before April would … be surprising. kalshi.com/markets/kxspcelau…

BREAKING: China's government has 'taken possession of potentially sensitive F-35 stealth aircraft parts' that were diverted to Hong Kong while being sent to the US from Australia, and 'so far has not returned them', according to Bloomberg report.

When several people talk at once, a transcript can get messy fast. Our new Nemotron 3 Diarization model tracks who spoke when, even when voices overlap. It handles up to eight speakers, has 100M parameters, and is now available on @huggingface 🤗

NEW As part of our ongoing efforts to crack down on illegal robocalls and text, the FCC has published step-by-step guides that enable consumers to turn on blocking and filtering tools. They are specific to your type of phone and carrier. Find them here: fcc.gov/blockit

A CS undergrad asked me where he could have most effect in the AI age. I said probably at either extreme: either close to the technology, actually making LLMs, or close to the customer, using AI to give them exactly what they want. Or maybe both if you can stretch that far.