Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. I Built Non-Autoregressive Decision Models with RL a Year Ago(convaiinnovations.com ↗)
    165comments
  2. AI-generated posters don’t have to be horrible(john.hartnup.uk ↗)
    555comments
  3. Human brain is two separate organs, Stanford Medicine-led research finds(stanford.edu ↗)
    197comments
  4. A graphical desktop for the ZX Spectrum(github.com/mindbox77 ↗)
    82comments
  5. Tin: full-text search for Postgres(planetscale.com ↗)
    56comments
  6. “The Secret Life of Circuits” is here(coredump.cx ↗)
    58comments
  7. Android 17 is the first since 3.x to add new APIs without releasing to the AOSP(grapheneos.social ↗)
    591comments
  8. Black Holes or Black Hole Stars? Astronomers Spar over 'Little Red Dots'(quantamagazine.org ↗)
    22comments
  9. Supabase (YC S20) Is Hiring for OrioleDB(supabase.link ↗)
    discuss
  10. New evidence for hidden chambers beyond Tutankhamun's tomb(nature.com ↗)
    13comments
  11. GPT-6 Astra Solves a WWI German Radio Cipher(prinzai.com ↗)
    144comments
  12. San Francisco Onion Futures Company(onionfutures.com ↗)
    133comments
  13. People who know the most often sound the least certain(vrash.substack.com ↗)
    7comments
  14. Almost Never Use AI to Write Anything Substantive(erichgrunewald.substack.com ↗)
    29comments
  15. Cloudflare Quick Tunnels(cloudflare.com ↗)
    302comments
  16. How to Write with an LLM(sockpuppet.org ↗)
    365comments
  17. If math is more than proof, we need to better celebrate the rest of it(terrytao.wordpress.com ↗)
    207comments
  18. What Zig felt like, coming from Rust(besok.github.io ↗)
    145comments
  19. Suzanne Ciani's Buchla Cookbook(echo.orpheusinstituut.be ↗)
    discuss
  20. Asking Authors About Their Own Papers(medium.com/tmlrorg ↗)
    50comments
  21. You can run Git on object storage if you re-make packfiles(tigrisdata.com ↗)
    29comments
  22. Saving another 100TB of RAM(cloudflare.com ↗)
    93comments
  23. Communication by means of modulated Johnson noise(pnas.org ↗)
    23comments
  24. SDCC – Small Device C Compiler(sourceforge.net ↗)
    26comments
  25. How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip(ieee.org ↗)
    123comments
  26. Adventures in Microcontroller Circuit Debugging(bigmessowires.com ↗)
    1comments
  27. Ray Ozzie and the Optimism of Being Early(reproof.app ↗)
    20comments
  28. Science Is Open Software(jepedersen.dk ↗)
    53comments
  29. OpenJev(openjev.com ↗)
    285comments
  30. Why building a Rust LSP is hard(rust-glancer.github.io ↗)
    52comments

GPT-5.6 Sol Pro solves open problem in convex optimization

26 pointsby 2mo agomedium.com
2 comments
2mo agoHN ↗

Discussion on Reddit: https://old.reddit.com/r/math/comments/1uxj3cy/after_openais...

Of note: the author had AI help synthesizing the actual 10-page prompt that resulted in the proof! A powerful tool when you know what you're doing.

Did you have first an AI help designing the prompt?

> yes, I did! I basically had 5.6 Sol synthesize existing closely related work and their approaches, the past ideas I had, with OpenAI's prompt that had a lot of the presumably important mechanisms for how exactly the agent should act. Especially from the "results that do not count" section onwards is a lot of input from Sol.

2mo agoHN ↗

For anyone dooming I’ll just leave you with this bit. It still takes quite a bit of knowledge to get it going:

After the release of GPT Sol 5.6, I used a very long prompt, 10 pages long in my paper