
🧠 Energy Minimalism | 🌲 @imperialcollege & AIAI | ex 🟥 @MiniMax_AI & 🟨 @tripoai & 🟦 @dify_ai |⚡️ Agentic AI | ✨ Views are my own
Joined April 2023
- Tweets3.5K
- Following1.2K
- Followers7.8K
- Likes17.5K
Lyson Ober retweeted
Many people think any given ML project is 99% training.
In reality, it’s 50% evaluation, 40% data cleaning, 8% integration, and 2% training.
The first two set the noise floor for learning. No ML magic matters; the model cannot lower the noise floor, as that’s the optimal bound of Shannon encoding of your data.
Thus, not a single day goes by without me thinking about ontology. Even the old labels have to be constantly reviewed.
Lyson Ober retweeted
AI SDK now supports agent harnesses like Claude Code, Codex, and Pi with sandboxed sessions and AI SDK-compatible streams:
𝚌𝚘𝚗𝚜𝚝 𝚊𝚐𝚎𝚗𝚝 = 𝚗𝚎𝚠 𝙷𝚊𝚛𝚗𝚎𝚜𝚜𝙰𝚐𝚎𝚗𝚝({
𝚑𝚊𝚛𝚗𝚎𝚜𝚜: 𝚌𝚕𝚊𝚞𝚍𝚎𝙲𝚘𝚍𝚎,
𝚜𝚊𝚗𝚍𝚋𝚘𝚡: 𝚌𝚛𝚎𝚊𝚝𝚎𝚅𝚎𝚛𝚌𝚎𝚕𝚂𝚊𝚗𝚍𝚋𝚘𝚡(),
});
Available in canary: 𝚗𝚙𝚖 𝚒 𝚊𝚒@𝚌𝚊𝚗𝚊𝚛𝚢. We welcome your feedback as we bring agent harness portability to the ecosystem, with excellent DX.
vercel.com/changelog/program…
Lyson Ober retweeted
没有壁垒的增长好比航空业:每年增长 5%,但一百年下来累积利润为零。
Chris Hohn did a 90-minute sit-down with Nicolai Tangen and then dropped an investor letter the FT got hold of last week.
You’d think the guy who printed a record $18.9B last year would be doing victory laps. Instead he’s quietly rewiring his whole portfolio.
My favorite takes from both:
1.The most important thing in investing isn’t growth. It’s barriers to entry. Growth without a moat is the airline industry: 5% volume growth for 100 years and basically zero cumulative profit.
2.There are only about 200 companies on earth he considers high-quality and investable. His fund holds 15.
3.Average holding period: 8 years. Some positions 13. “You have to hold the company forever, because the stock market may be at very bad prices when you want to sell.”
4.His real test for a moat: can the company price above inflation? A 20% margin business that prices 1% above inflation grows profits 5% faster than revenue. Forever. Almost no companies can do this.
5. Industries he won’t touch: banks, autos, retail, insurance, tobacco, asset managers, fossil fuel utilities, airlines, wireless telecom, media, advertising. On banks: “sooner or later someone without a lot of intelligence comes to run them, and then it can be toxic.”
6.On AI generally: call centers go bankrupt. Indian outsourcing coders are next. But for everyone else, AI lowers costs and raises productivity. Companies with real moats become MORE valuable.
7. Here’s the punchline. The FT got hold of his investor letter. He cut his Microsoft stake from 10% of the fund to 1%. Roughly $8B sold. He’d held it since 2017 through a 400% rally. His reason: AI could disrupt Office and Azure faster than the market thinks.
8.He moved that capital into Alphabet. Doubled it from 3% to 5%. Now his largest tech position. The world’s best quality investor sold Microsoft and bought Google because he thinks Google’s moat is more durable in an AI world. Not the consensus trade.
9.The underlying thesis: “AI eats software.” If AI agents do the work humans used to pay per-seat SaaS licenses for, the whole SaaS model gets re-rated. Oracle, Adobe, Salesforce all ~40% off highs. Microsoft 25% off. Market is starting to agree.
10.When to sell? Not when something gets expensive. When conviction drops. Valuation is one variable, conviction is the other. What kills you isn’t being wrong, it’s permanent loss of capital.
11.He admits hardcore activism doesn’t work anymore. Too much of the shareholder base is passive index funds. And even when activism wins, you usually win in a bad business. “The business always wins.”
12.Counterintuitive take: there are more good companies in public markets than in private equity. The best businesses are too big for PE to buy. And when public companies sell something to PE, they’re selling the assets they want to get rid of.
13.On intuition: “thinking without thinking.” Pattern recognition from 20 years of reps. It’s how he sniffed out Wirecard while the German establishment was defending it. “Most investors trust authority too much.”
14.He basically stopped shorting. “You’re going to be eventually right but not be able to fund the losses.” The first guy to short Wirecard had to cover 19 years before it hit zero. Buffett told him he and Charlie studied shorting and concluded it was too hard.
15.He gives almost everything away. ~$500M a year. $10 prevents an unwanted pregnancy in Africa. $40 saves a child from severe malnutrition. $50 prevents permanent blindness.
16.Tangen asks: advice to young people? Hohn, who runs the world’s most profitable hedge fund: “Go on a spiritual path.” The guy who made $18.9B last year ends the interview saying only purpose and meaning matter.
The headline: the world’s best quality investor just sold his biggest tech compounder because he thinks AI is breaking the moat. Quietly, with conviction, on an 8-year horizon, while everyone else is still buying the AI winners of 2023.
Lyson Ober retweeted
Thanks to the community report, we recently identified a PR github.com/vllm-project/vllm… that attempted to solve a non-existent issue and was submitted as part of a “PR training” workflow for resume building.
The contributor involved has been banned from the vLLM community.
This kind of low-signal contribution increases maintainer review overhead and creates unnecessary operational costs for open-source projects.
As AI coding agents make generating large volumes of small PRs increasingly cheap, open-source communities will need to explore new ways to preserve contribution quality and reviewer trust.
While we are investigating how to deal with AI slop, we continue to highly value contributions from real users solving real production problems.
If you have an important contribution that has not yet received maintainer attention, please email us at:
[email protected]
Using a verifiable company or university email, include:
- your production or research use case
- the problem you encountered
- how your contribution addresses it
This helps us better prioritize impactful contributions while keeping the vLLM community open and collaborative.
As AI makes virtual contributors look increasingly real, authentic human collaboration matters more than ever.
vLLM’s mission remains unchanged: to make LLM inference easy, fast, and cheap for everyone — and we will continue working toward that goal.
Lyson Ober retweeted
🎤 Keynote announcement:
@trq212 (Thariq Shihipar), Member of Technical Staff on Claude Code at @AnthropicAI, is keynoting #CAIS2026.
Thariq's "Lessons from Building Claude Code" series on Skills, prompt caching, tool design, and "unhobbling" is required reading for anyone building agentic systems.
We're thrilled to have him.
📍 San Jose · May 26–29
🔗 caisconf.org
Lyson Ober retweeted
创业和在打工的最大区别就是,风险与兜底。
在大厂,相对来讲是低风险的,因为有平台帮你兜底,有创始人帮你兜底。你做失败了,大不了换一份工作。
创业是没有人帮你兜底的。(如果你有岳父帮你兜底,当我没说),创业的人,必须要敏感,对数字敏感,对成本敏感,对增长敏感,对风险敏感。
有兜底很多时候,就会丧失敏感性。
创始人,对自己诚实,是基本素质。
绝对不能骗自己,犯错了,就要挨打立正。
市场不会错,错的只是你自己。
David很诚实呀,这样的创始人很有魅力
David说的没错,其实在创业的过程中,发现赚钱的秘诀就是创始人本身的认知。创始人本身认知高,其实赚钱就很轻松。创始人本身认知低,其实就很难赚钱
但是在创业的过程中,没有人是先天就完美的,就像David这样的,他说他选择了TPU,而不是英伟达的显卡。他诚实地说,这不是Google骗了他,是他自己骗了自己
真的非常真诚呐,这就是创业最真实的写照,创业除了提升认知以外,最重要的第二点就是不要自己骗自己
很多创始人在创业的过程中,为了快都会走捷径,都会投机。然后投机的过程中就会自己骗自己,自己想象自己的未来
其实创业就是提高认知,不要投机,坚持下去就一定会成功的
David这样诚实的创业者非常难得,而且他还把他创业的过程中分享出来了,给他点赞
Lyson Ober retweeted
📣Meet Qwen3.7-Max — our latest flagship, made for the Agent Era.
A versatile foundation for agents that actually get things done:
🧑💻 Coding agent, end to end. Frontend prototypes, multi-file refactors, real debugging — nails it.
🗂️ A reliable office and productivity assistant. Get your work done through MCP integrations and multi-agent orchestration.
⏱️ Long-horizon autonomy. 35 hours straight on a kernel optimization task — 1,000+ tool calls, zero hand-holding.
🔌 Scaffold-agnostic. Claude Code, OpenClaw, Qwen Code, or your own stack. Consistent reliability everywhere.
API's up on Alibaba Model Studio. You can also take it for a spin on Qwen Studio.
Go build something wild!🏃🏃♂️
📖 Blog: qwen.ai/blog?id=qwen3.7
✅ Qwen Studio: chat.qwen.ai/?models=qwen3.7…
⚡️ API:modelstudio.console.alibabac…
Lyson Ober retweeted
一个我熟知的独立开发者,因长期熬夜工作,前几日猝死。
震惊与悲痛之余,这件事情再次给我们所有人警告:
1. 身体,是一次性的
前一秒看似正常运行,下一秒可能直接宕机
2. 钱,是赚不完的
时间很晚了,手上还有工作,只要世界不爆炸,那就明天做
3. 熬夜,但保证充足睡眠
你不能指望每天 3 点睡,第二天 8 点起床,却没有任何代价一样






















