My First Week with GPT-6 Astra(www.vincentschmalbach.com)
4 pointsby vincent_s4 hours ago1 comment
  • jdw644 hours ago
    Using Astra, I get the feeling that LLMs have hit a ceiling. To be exact, it's not necessarily a hard limit, but a point of diminishing returns where squeezing out tiny performance gains yields barely noticeable improvements in practice.

    I switched from GPT-5.6 SOL to Astra, and while it is genuinely impressive at generating 3D meshes and similar assets, its coding capabilities actually seem worse than SOL in many cases. Granted, this is just my subjective impression.

    There is a massive gap in token pricing between GPT-5.2 and GPT-6 Astra, yet the actual leap in quality feels negligible.

    Will we really see a massive performance spike from where we are now? I'm somewhat skeptical, and I wonder what actual AI researchers think about this.

    • vincent_san hour ago
      I have the exact same feeling. For me the last big leap in models was when the Codex-Family came out (GPT-5.1-Codex to GPT-5.3-Codex) and then the GPT-5.4/GPT-5.5 that brought the "Codex" capabilities back to the main stream. The GPT-5.6 family feels very different, might be a little bit more intelligent but behaving very differently, and GPT-6-Astra is along those lines. That's why my main driver still is GPT-5.5.
    • Bluestein3 hours ago
      I feel same, and think this is perhaps why the rushed "AGI" call was made.-