Work in progress: 181,232 of 181,233 sites checked so far, more every week · updated 2 Nov 2026Missing a blog? Tell us

Top all-time · #23928

K

Kanta Hayashi

Personal blog · kantahayashiai.github.io/posts · on HN as kantahayashiaimath

Kanta Hayashi's personal blog at kantahayashiai.github.io covers ai and math. Since 2026, 3 different Hacker News readers have linked it 4 times, most often “Jev Does Not Play Dice: 83% probability, 19% accuracy on a hidden fair die roll”. It is still writing: the last post was on 2026-09-19, and you can follow it by feed.

Visit the blog
4mentions?
3HN people?
2026–2026first – last linked?

Too few sampled comments (1) to say how often HN recommends it.?

Why HN links it

Linked in a thread on how calibrated Jev's outputs are, using a dice example where the expected output is a probability distribution rather than random generation.

Summarised by Claude from a sample of 1 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.

Thinking about LLMs.

The author on HN?

kantahayashi · 63 karma · member since 2026

Kanta Hayashi. Student at the University of Tokyo, working on LLM research. Blog: https://kantahayashiai.github.io GitHub: https://github.com/KantaHayashiAI X: https://x.com/KantaHayashiAI

From their public HN profile, which links to this site.

Most-linked posts?

  1. Jev Does Not Play Dice: 83% probability, 19% accuracy on a hidden fair die roll
    3 people · 4 mentions · first linked 2026-09-23

What Hacker News says?

  1. I tested Jev with a fair die 400 times without telling it the die result. The true probability of face 1 is 1/6, but Jev always chose face 1 and the probability it returned was about 83%. I also tested with a fair coin 200 times and got 0.92 probability. I did several tests and I think Jev is good at problems with a correct answer but weak at problems about actual probabilities whose answers can't be known at all. W…

    Comment linking /posts/jev-does-not-play-dice/

    on “Jev Can't Be Calibrated”
  2. Submitted Jev Does Not Play Dice: 83% probability, 19% accuracy on a hidden fair die roll as a story · 4 points

    HN discussion
  3. In the dice example[0] and in this one, the expected output is a distribution e.g. "1: 16.7%, 2: 16.7%...", not a random generation. I agree that in implementation , the architecture is not designed to accurately calculate or incorporate any known uncertainties like these, but the task itself is completely reasonable and arguably trivial. IMO the lesson here is: even in trivial cases, Jev's outputs are just ~reasona…

    Comment linking /posts/jev-does-not-play-dice/

    on “How accurately calibrated is Jev?”
  4. Submitted Jev does not play dice as a story · 2 points

    HN discussion

Search all of HN for kantahayashiai.github.io