Live data from Hacker News

Viewing profile — dippogriff

dippogriff

HN member
Joined
Mon, Oct 27, 2025, 1:35 AM UTC
HN karma
2
Public activity
13 items

About dippogriff

No profile information was provided.

Recent public activity

  1. comment
    Comment #48777563

    I want to see more text-free foundation models

  2. comment
  3. comment
    Comment #48687154

    If the labs weren't so aggressive with building datacenters in people's backyards, this could've been a different story. People don't like it when pipelines are built in their back…

  4. comment
    Comment #48687048

    They tried that a few times and the mistakes have had consequences.

  5. comment
    Comment #48686988

    This is excellent! Very useful takeaways. Being able to properly do continuous training in production is key with robotics data being so hard to come by.

  6. comment
    Comment #48686807

    Great work showing on how brittle these GUI benchmarks can be! Love the visuals. I wonder if SFT is the problem here as opposed to the coordinate discretization; what happens with …

  7. comment
    Comment #48679075

    This is cool. Creative ways to do external verification is the only path to solving training on LLM slop

  8. comment
    Comment #48670082

    Neat! minor nit - would be nice if the esc button took you back to the list, instead of having the click the X button

  9. comment
    Comment #48663715

    The current way benchmarks are done and are accepted by the community makes for really uninspired work. Until we're willing to break out of this rigid evaluation format prone to cr…

  10. comment
    Comment #48663384

    Edge models will get much better after the current insane capex and organic data for pre-training is dried out. But hard to see how the best open source models will ever come close…

  11. comment
  12. comment
    Comment #48655693

    Vice versa, the exaggeration of vitamin D is mildly worthless. Some need supplements, some don't.

  13. comment
    Comment #48655664

    I'm a fan of this direction. For me the most interesting use case for these world models isn't even training, it's verification. If this thing or some idealized version of it can a…