Live data from Hacker News

Viewing profile — JacobAsmuth

JacobAsmuth

HN member
Joined
Tue, Nov 18, 2025, 10:18 PM UTC
HN karma
113
Public activity
90 items

About JacobAsmuth

No profile information was provided.

Recent public activity

  1. comment
    Comment #49236375

    I would prefer that Trump keep all the power personally. He's done such great work recently I don't see why he shouldn't be in control of the machine god

  2. comment
    Comment #49190636

    You're hallucinating facts again.

  3. comment
    Comment #49190623

    I've said this for a long time. This inability of AIs to generate novel explanatory hypotheses is a big blocker for their ability to automate jobs like accountant, middle manager, …

  4. comment
    Comment #49186417

    The irony of using Claude to write this is almost to much to "hold". (Claude loves to tell you to hold things in its writing). Here's another great Claudism: "The replies had a dif…

  5. comment
    Comment #49186295

    Really? You prefer not to see the top end of the distribution? Why?

  6. comment
    Comment #49185858

    > mobile era CEO He was the guy branding Google an AI-first company back when they invented the transformer.

  7. comment
    Comment #49185836

    So then it is similarly telling that compute is second on the list? They have more talent than compute? Layoffs incoming?

  8. comment
    Comment #49173801

    Dallonians

  9. comment
    Comment #49173758

    only 30 upvotes in 3 hours but 50 comments(many negative) is not a great signal.

  10. comment
    Comment #49173488

    Me after I spend a day at the beach: #BA3A17

  11. comment
    Comment #49148482

    Why do you think you've prevented third party supply train risk when you're training on a 30+ Terabyte blob that you have little understanding of its contents. It's essentially ent…

  12. comment
    Comment #49148476

    Can you give an example of a question that you would google, but your colleagues would "Ask Claude" and they get a hallucination but you get the right answer?

  13. comment
    Comment #49074894

    1e21 flops is hilariously wrong. for reference the llama 3 8B model ( https://arxiv.org/pdf/2407.21783 ) used 10 times that many flops. This model is 350x bigger in total params an…

  14. comment
    Comment #49074850

    Now you're thinking like management!

  15. comment
    Comment #49074840

    Why do you feel that creating the LLM from scratch is sufficient to trust it? Are you suggesting that you personally would read all 15 trillion tokens (plus every single agentic tr…

  16. comment
    Comment #49073601

    It's 3/15 - https://openrouter.ai/moonshotai/kimi-k3 If you're going to open source your model, why would you set your own price high enough that other providers could easily and p…

  17. comment
    Comment #49042582

    Their competitors aren't able to serve models as good as Gemini 3.6 Flash to billions of users

  18. comment
    Comment #49042574

    buddy man

  19. comment
    Comment #49042569

    You're confused by this? Fable > Opus > Sonnet > Haiku. Sort by price. Most expensive = better. Are you an alien? lol

  20. comment
    Comment #49042566

    All providers are equally trustworthy :)

  21. comment
    Comment #49042562

    AA isn't the best way to measure relative cost in real world use because some of those benchmark questions are extremely hard for the models. Some models give up quickly on hard qu…

  22. comment
    Comment #49042261

    What? Claude plays Pokemon one-shots the entire game without any harness other than claude code and game screenshots

  23. comment
  24. comment
    Comment #48996164

    Marco econ 101 disagrees :) Ask your favorite LLM to explain the "yield curve" to you

  25. comment
    Comment #48995795

    How has your experience been with Gemma 4?