We've dropped our technical report of "AI-Researcher "for "Agentic Scientific Discovery" - a fully automated system that transforms scientific innovation with LLM agents. 🎉 Major Release: AI-Researcher Update 🚀 🔗 GitHub: github.com/HKUDS/AI-Resea… 📄 arXiv:

If the placebo effect is real (and real even when you know it is just a placebo effect), then shouldn’t you to try believe you’re eating the best food, meeting the best people, living in the best place, working the best job, etc., just to make yourself happier?

💖Cool! And we will improve the performance of Multi-image with qwen-image-edit in the next version.

This is the real deal: „But the new analysis suggests that energy estimates are currently a moving target, as the company says its data shows the energy drain of a search has dropped by a factor of 33 in just the past year.“ Energy will be the limiting factor in a few years.

Being unoffendable is a massive advantage in today's world. The ability to let nonsense or a differing opinion pass without a reaction will change your life. No response, no letting it get under your skin, and definitely no letting it alter the course of your day.

This guy literally shows you how to read so you never forget

Not against women working I just don’t think they should have to. Just my opinion I don’t want any trouble

A lot of knowledge work is compressible. Repetition, standard formats, and quick feedback mean large swaths of what knowledge workers can be automated. Can the spec be expressed in language? Are there many prior examples? Is correctness cheaply checkable? If yes, LLMs excel. If

XBOW finds GPT-5 to be much better at hacking than OpenAI finds it to be, which goes to show that scaffolding matters a lot! And differences in scaffolding can really complicate AI evaluations!

Start planning now to go to mass tomorrow. They have them all day and you’ll feel really nice after. Masstimes.org

One of the most brutally effective ways I know of becoming successful is to actually try really really hard.

Good chart from Gavekal, making the point that a solid share of AI capex (especially chips) in the US is imported

Outreach is the fastest shortcut to revenue.

“What if they say no?” What if they say yes?

The top open source LLMs have some interesting differences in architecture and training methods. I read all of the papers in depth to break them down in this video (and my debut on the YC YouTube 😅) Check it out and let me know what you think!

Lebubus are obviously satanic and the children and redditors exposed to them will have to be put down to protect the herd. Obviously

I'm learning the true Hanlon's razor is: never attribute to malice or incompetence that which is best explained by someone being a bit overstretched but intending to get around to it as soon as they possibly can.

You should never smoke, and you should never ride a motorcycle

Fear is a PSYOP