Rasmus Ros

Hi, I'm Rasmus. Head of engineering by trade, building Eignex on the side: open-source infrastructure for optimization loops in production. Posting here as I build it. eignex.com/posts/building… #buildinpublic #opensource

Good HN blog article on using LLMs at a slower pace: using multiple agents on the same PR to surface issues, then reviewing and filtering the comments yourself, is a pretty sensible way to trade speed for better code. nolanlawson.com/2026/05/25/usi…

@JackFriedson They always take the most obvious iterative step and never take the step back you need for good architecture.

another day of wondering whether the efficiency gains from coding agents really justify the amount of time I spend undoing their bullshit design choices

@jonathan_wilke @cursor_ai That is a good idea. I use plan a lot, and one of the conflated issues is probably trying to make too big a change at once.

I think @cursor_ai and other AI coding tools should default to a “/grill-me” planning mode that asks way more questions before writing code. One of the biggest weaknesses of current AI models is that they don’t challenge assumptions enough. Please ask me more!

@asmah2107 Pretty much. If the answer can be checked quickly, the system gets better fast. If you need a human, a lab, or six months in the real world, progress slows down a lot.

AI improves exponentially in domains where its work can be instantly and objectively checked. If it takes a human, a lab experiment, or months of real-world testing to verify if the AI's answer is actually good, that field is much harder to automate.

@HoytEmerson Right. Don’t optimize for scale until you need it.

The boring version of your data pipeline is often the best one. Parquet → S3 → DuckDB → BI tool. No distributed compute, no DAG framework, no orchestration complexity. It works reliably until it doesn’t, then you actually know why.

@cnakazawa It’s the other way around. If the Rust Bun has now is poor quality and full of unsafe, that’s not going to help adoption.

People are underestimating how big of a deal Bun’s Rust rewrite is for the project. I believe it’s existential, and it’s going to be huge for adoption. Personally, the instability and memory issues made me wary, and the rewrite makes me more likely to use Bun.

@haider1 Yeah im awake. What am i supposed to do again?

the computers are speaking. wake up. wow so much is happening internally at the big labs -- and the fact that this is coming from a former openai researcher makes it even harder to ignore

@8teAPi The conversation is driven by those who speak up. No way around it.

Given poster/lurker ratio of 1/99 I often wonder if early AGI gets defined purely by the posters.

@tunguz Selling hand crafted artisanal C-functions $20. Custom bespoke assembly $50. Every line written on paper from local trees.

The future of humanity is artisanal.

@sunnyyuych Efficiency in speed is one thing. When people get hooked on this technology it's too easy to shut your brain down and reach for it instead of searching or reading a book. I wonder tho if this is also an illusion or do you actually use less energy from just using your AI?

“AI, what color do I get from mixing black and white?” Why do people turn to AI for simple tasks that they could easily do themselves? In our new preprint (also to appear at CogSci 2026!), we investigate the mechanisms and dangers of people over-using AI on easy tasks.

@JosephJacks_ The world is split in two atm: 1) people talking about AI and 2) people talking about people talking about AI.

Absolutely spectacular things are possible. And people are so distracted with LLMs. It’s really baffling.

@trikcode $20/month is cheap compared to hosting it yourself. And opencode is not as good as Claude code. It's good that we have the option but it's not a rational choice for individuals. If you're a company that changes the equation.

DeepSeek is free and almost as good as Claude. but we still pay $20/month because we've convinced ourselves the paid one understands us better.

@__paleologo You have to give him credit for putting his balls on the line with specific predictions in "Deep Learning Is Hitting a Wall," even if most of them were wrong.

Reading old Gary Marcus essays, for contrarian views, because there was some debate on X. newyorker.com/news/news-desk… nautil.us/deep-learning-… garymarcus.substack.com/p/misplaced-pa… But, the more I read, the less interesting it is. So I am stopping and going back to work.

@AlexanderKalian A lot of AI can be both real demand and bad equity math. If serving the demand needs huge capex, fast model depreciation, and everyone cuts price at once, users win and investors still eat it.

"Cool tech" and "good investment" are not always the same thing. For example, a Moon base is cool tech - but a company trying to raise $100B to build one is likely not a good investment. Whether or not AI is in a market bubble, entirely hinges on the "good investment?" question - i.e. if companies like OpenAI and Anthropic are valued reasonably in terms of business fundamentals, risk-reward, and comparison to the wider market. OpenAI for example, is currently unprofitable and valued at ~34x revenue. Anthropic is likewise also deeply unprofitable, yet valued at ~21x revenue. This is at a time when smaller and more profitable AI start-ups are lucky to get valued at 7x revenue by PE firms. This is also at a time where big tech giants (e.g. Alphabet, Meta, Amazon) are typically valued at 21x-31x *profits* - not revenue. In terms of revenue multiples, they are typically valued at 3x-11x. This is also at a time where companies in areas such as biotech, entertainment, e-commerce etc. are trading at historically low valuations compared to business fundamentals. So yes, AI does appear to be in a market bubble. Pure-play AI companies are significantly overvalued.

@relizarov The question is if it's worse for mathematicians? SWEs at least have the skills to develop or use AI. Pure mathematicians will have to rely on AI tools developed by others if they want to use it in their work.

Mathematicians are now having their “AI moment,” asking the same uneasy questions about the future of their profession that software developers asked first.

@madiator Everyone needs MCP. MCP is dead.

Software is eating the world. AI is eating the world Attention is all you need RAG this, RAG that. Agentic this, Agentic that. Context engineering is what you need. RAG is dead. Long live RAG. Harness engineering is what you need. Harness is the backend. What did I miss?

@codetaur Half the value is making yourself write down the tradeoffs in one place. Then the answer is usually obvious. It's just a better rubber duck.

my typical input to codex is like: "okay I want to achieve X. but I can think of like 4 ways of doing it. if we do Y we have to deal with Z. if we do A we have to deal with B. I think? maybe? idk what I'm talking about. is there some way to do C without D? what do you think?"

@teortaxesTex That's not an honest read of the situation. More like Google is the only one offering a scarily broad ecosystem and product stack.

Interesting how different are frontier labs' notions of progress towards AGI. OpenAI: "we've disproven an old conjecture in math" Anthropic: "we've discovered ALL the vulnerabilities" DeepSeek: "we've made context free" Google DeepMind: "we've reduced the batch size for Flash"

I'm still thinking about the result by OpenAI solving a real conjecture. How much of it is LLMs capability of being good at everything vs doing something really inspired? Or that they probably tried pretty much all major unsolved problems and got lucky on one? Or maybe I'm coping 🙃

@rohanpaul_ai I question this narrative. Calling AI a war with China mostly just makes monopoly, surveillance, and labor displacement sound necessary.

Vinod Khosla: "We are in a techno-economic war with China, and we shouldn't call it anything other than a war. Whoever wins this AI race will win the economic race and will win the race for socio-economic power and influence globally." ~ Vinod Khosla, Co-founder of Sun Microsystems & Billionaire Silicon Valley venture capitalist. --- From 'Fortune Magazine' YT Channel