Kevin Schawinski
@kevinschawinski
Astrophysicist Turned AI Entrepreneur | CEO at Modulos AG | Bridging Science and Responsible AI | Yelling at Claudes
You might like
Opus 4.5 doesn't feel like a "4.5" version of 4.1. It feels like... something distilled from an even more powerful model? Anthropic is cooking.
👀
👀👀👀👀👀
The EU Parliament's JURI committee will vote next week on whether to challenge the Commission’s decision to withdraw the AI Liability Directive before the EU Court of Justice - in a move that could replicate the path followed by the SEP proposal. mlex.com/mlex/articles/…
It's a deepseek fine tune
Today, we are releasing the best open-weight LLM by a US company: Cogito v2.1 671B. On most industry benchmarks and our internal evals, the model performs competitively with frontier closed and open models, while being ahead of any US open model (such as the best versions of…
Can’t wait to show you guys what I’m building with Interintellect AI 😍 Really nothing like this in the market & we get to test it 1-2x/wk in Interintellect (in the semi-wild) which is huge. Next demos: Sunday, Wednesday HIRING: have open opportunities for character engineers!
I notice that Gemini 3 in Antigravity is worse than Sonnet 4.5 in following some instructions....
I guess the principle of legal certainty means nothing to the EU commission. "We're moving the deadline, but maybe we'll move it back, who knows? YOLO" Absolute bananas and making the Commission look ridiculous.
BREAKING: The Commission’s digital omnibus on AI proposes delaying the AI Act’s high-risk rules under Annex III to Dec 2027 and Annex I to Aug 2028. But the Commission could anticipate the deadline, giving a six-month or 12-month notice, respectively. digital-strategy.ec.europa.eu/en/library/dig…
Now this looks interesting. Getting errors because there's not enough capacity right now....
Has been driving me crazy since the ChatGPT moment. Most people “hallucinate” most of their answers all day long.
It's interesting how people hold AI models to a higher standard than people. Eg try asking people 2-4 questions in one go -- they tend to answer ~60% of them (and that's for educated people!). A 60% score on that eval would be considered a crippling deficiency for a model!
We built our whole society around the idea that you need expertise to write things (in this case, planning objections) and this is no longer true for $20/month.
Future generations will rightfully condemn us for the immeasurable cruelty we visit upon animals in animal agriculture today.
The pig area at the Royal Ag Fair shows a mother pig in a cage so small she can barely move. Approximately nobody cares, and she is not hidden (she is a prominent part of their display).
Did... did the UK government miss a zero there with that salary?
The style difference between Claude code and Openai codex is striking. Can you imagine Claude answering like this?
👀
We will introduce “made in Europe” criteria for public procurement in certain strategic sectors. Ensure that new foreign investments in our industry are truly in Europe’s interest. And step up support for strategic industries, like cars and batteries
United States Trends
- 1. Lakers 50.1K posts
- 2. Dillon Brooks 7,036 posts
- 3. Kanata 11.1K posts
- 4. Bron 24.9K posts
- 5. Giants 86K posts
- 6. #WWERaw 71.4K posts
- 7. Patriots 132K posts
- 8. Dart 36.2K posts
- 9. Suns 19.2K posts
- 10. Collin Gillespie 2,049 posts
- 11. #AvatarFireAndAsh 3,063 posts
- 12. Drake Maye 24.4K posts
- 13. STEAK 10.6K posts
- 14. Ryan Nembhard 4,253 posts
- 15. Diaz 34.2K posts
- 16. James Cameron 4,691 posts
- 17. Devin Williams 7,757 posts
- 18. Gunther 15.3K posts
- 19. Pats 16.3K posts
- 20. Devin Booker 2,640 posts
Something went wrong.
Something went wrong.