• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Suhail Doshi Disappointed After One-Shot Testing with /goal

    Playground AI founder shares his test results using single prompts on an AI tool.

    MA
    SU
    RG
    6 Sources, 24d ago, first seen 24d ago

    TLDR

    Suhail Doshi replied to investor Martin Casado on X that he routinely one-shots tasks with the /goal command to test AI model capability jumps. Doshi stated he was fairly disappointed. The exchange, which included replies from Dax Raad and Bilawal Sidhu, centered on prompting methods, usability improvements, and persistent model limitations. Bilawal Sidhu noted the approach works for cleaning existing codebases. Martín Casado observed that ergonomics improve but core issues show little progress. The posts contain no details on specific models, tasks, or outcomes beyond the stated reaction.

    Combined views

    39.2K

    6 Sources, first seen 24d ago

    Combined views

    39.2K

    6 Sources, first seen 24d ago

    207 likes
    207 likes
    8 comments
    26 saves
    2 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    8 comments
    26 saves
    2 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    6 Sources

    @Suhail@martin_casado Nah, not a skill issue. I always try one-shotting things with /goal now to see if it jumps in capability and I'd say I was fairly disappointed.
    @bilawalsidhu@martin_casado It’s really good as deslopping existing code bases - give that a shot!
    @thdxr@martin_casado i've been thinking more in terms of usability instead capability for a bit i think the models continue to get easier and more fun to use which makes me feel less frustration and use them more but yeah i think my workflow is finally settling
    @martin_casado@thdxr Right. Ergonomics keep getting better. But I have a set of outstanding issues I throw all new models at. And that hasn’t demonstrated much of an improvement.
    @RyanGreenblatt@martin_casado Astra looks pretty similar in coding benchmarks to 5.1 Fable while being a big jump in some other tasks. So not too surprising?

    6 Sources

    @Suhail@martin_casado Nah, not a skill issue. I always try one-shotting things with /goal now to see if it jumps in capability and I'd say I was fairly disappointed.
    @bilawalsidhu@martin_casado It’s really good as deslopping existing code bases - give that a shot!
    @thdxr@martin_casado i've been thinking more in terms of usability instead capability for a bit i think the models continue to get easier and more fun to use which makes me feel less frustration and use them more but yeah i think my workflow is finally settling
    @martin_casado@thdxr Right. Ergonomics keep getting better. But I have a set of outstanding issues I throw all new models at. And that hasn’t demonstrated much of an improvement.
    @RyanGreenblatt@martin_casado Astra looks pretty similar in coding benchmarks to 5.1 Fable while being a big jump in some other tasks. So not too surprising?