aneryu

6 ms·

1. 1.▲
   by aneryu 28d ago

   I don’t think it’s that Sol loves to cheat; rather, every GPT model since 5.4 seems to have had that tendency.
2. 2.▲
   by aneryu 29d ago

   Based on my experience so far, code written by today’s frontier LLM is generally trustworthy, though it still requires human oversight.
3. 3.▲
   by aneryu 29d ago

   It would be even better if a version available to individual users were released soon.