conv.

All stories
AIQuiet 10d · day 18

New papers push AI recursive self-improvement beyond math and code benchmarks

A cluster of research papers on AI systems that improve their own training pipeline surfaced on Hacker News within a single day.

What to know

  • Multiple papers on AI 'recursive self-improvement' (RSI) reached Hacker News within roughly ten hours, suggesting a concentrated wave of research interest rather than one isolated finding.
  • The most detailed proposal, MetaRSI-v1, combines three types of self-improvement operators (data, scaffold, model weights) but is only validated on standard coding and closed-form science benchmarks, not open-ended real-world tasks.
  • The papers argue existing RSI work has been confined to machine-checkable domains like math and code, and call for extending it to broader scientific and engineering work where correctness is harder to verify.
  • Evidence available is limited to paper abstracts and bare HN submission listings, with no substantive comment discussion captured.

Zihan Tan Co-lead author, MetaRSI-v1 paperLeixin Sun Co-lead author, MetaRSI-v1 paperGuancheng Wan Author, MetaRSI-v1 paper

How it unfolded 5 developments, newest first · click a bar or a number to jump posts

Peak 8 pieces in 4h at Sep 16, 11 AM; 27 pieces over 18 days (5 articles · 9 posts · 13 comments) Sep 9, 11 PM — 1 piece · 1 article — Newswires 1Sep 10, 3 AM — quietSep 10, 7 AM — quietSep 10, 11 AM — quietSep 10, 3 PM — quietSep 10, 7 PM — quietSep 10, 11 PM — quietSep 11, 3 AM — quietSep 11, 7 AM — quietSep 11, 11 AM — quietSep 11, 3 PM — quietSep 11, 7 PM — quietSep 11, 11 PM — quietSep 12, 3 AM — quietSep 12, 7 AM — quietSep 12, 11 AM — quietSep 12, 3 PM — quietSep 12, 7 PM — quietSep 12, 11 PM — quietSep 13, 3 AM — quietSep 13, 7 AM — quietSep 13, 11 AM — quietSep 13, 3 PM — 2 pieces · 2 posts — Hacker News 2Sep 13, 7 PM — quietSep 13, 11 PM — 1 piece · 1 post — Hacker News 1Sep 14, 3 AM — 2 pieces · 2 posts — Hacker News 2Sep 14, 7 AM — quietSep 14, 11 AM — quietSep 14, 3 PM — quietSep 14, 7 PM — quietSep 14, 11 PM — 2 pieces · 2 articles — Newswires 2Sep 15, 3 AM — quietSep 15, 7 AM — quietSep 15, 11 AM — quietSep 15, 3 PM — quietSep 15, 7 PM — quietSep 15, 11 PM — 1 piece · 1 article — Newswires 1Sep 16, 3 AM — quietSep 16, 7 AM — 7 pieces · 1 article · 4 posts · 2 comments — Hacker News 5, Newswires 1, Mastodon 1Sep 16, 11 AM — 8 pieces · 8 comments — Hacker News 8Sep 16, 3 PM — 2 pieces · 2 comments — Hacker News 2Sep 16, 7 PM — quietSep 16, 11 PM — quietSep 17, 3 AM — quietSep 17, 7 AM — quietSep 17, 11 AM — quietSep 17, 3 PM — quietSep 17, 7 PM — 1 piece · 1 comment — Hacker News 1Sep 17, 11 PM — quietSep 18, 3 AM — quietSep 18, 7 AM — quietSep 18, 11 AM — quietSep 18, 3 PM — quietSep 18, 7 PM — quietSep 18, 11 PM — quietSep 19, 3 AM — quietSep 19, 7 AM — quietSep 19, 11 AM — quietSep 19, 3 PM — quietSep 19, 7 PM — quietSep 19, 11 PM — quietSep 20, 3 AM — quietSep 20, 7 AM — quietSep 20, 11 AM — quietSep 20, 3 PM — quietSep 20, 7 PM — quietSep 20, 11 PM — quietSep 21, 3 AM — quietSep 21, 7 AM — quietSep 21, 11 AM — quietSep 21, 3 PM — quietSep 21, 7 PM — quietSep 21, 11 PM — quietSep 22, 3 AM — quietSep 22, 7 AM — quietSep 22, 11 AM — quietSep 22, 3 PM — quietSep 22, 7 PM — quietSep 22, 11 PM — quietSep 23, 3 AM — quietSep 23, 7 AM — quietSep 23, 11 AM — quietSep 23, 3 PM — quietSep 23, 7 PM — quietSep 23, 11 PM — quietSep 24, 3 AM — quietSep 24, 7 AM — quietSep 24, 11 AM — quietSep 24, 3 PM — quietSep 24, 7 PM — quietSep 24, 11 PM — quietSep 25, 3 AM — quietSep 25, 7 AM — quietSep 25, 11 AM — quietSep 25, 3 PM — quietSep 25, 7 PM — quietSep 25, 11 PM — quietSep 26, 3 AM — quietSep 26, 7 AM — quietSep 26, 11 AM — quietSep 26, 3 PM — quietSep 26, 7 PM — quietSep 26, 11 PM — quietYesterday, 3 AM — quietYesterday, 7 AM — quietYesterday, 11 AM — quietYesterday, 3 PM — quietYesterday, 7 PM — quietYesterday, 11 PM — quiet 1–5
Sep 11Sep 13Sep 15Sep 17Sep 19Sep 21Sep 23Sep 25now · 3:14 AM ET
  1. 5

    Curated 'awesome-rsi' GitHub list posted to HN

    A GitHub repository collecting recursive self-improvement resources, 'awesome-rsi,' is shared on Hacker News shortly after the paper resubmission.

    • Perhaps I'm not understanding it correctly, but here's my take on what the paper is doing.Imagine you have a problem you want to solve (let's say, identify an OCR'd handwritten character, e.g. the MNIST Dataset). You tell 3 agents "Hey, each of you take a stab at getting really good at recognizing characters from this dataset. You can take 10…

      eggbrainHacker News11d agoview on Hacker News ↗
    2 more of the top 3 · 13 posts in this stretch
    • FYI; the paper is clearly a reference to Danijar Hafner's 'Dreamer' line of work, which was published in 2019, and which Danijar has continued to iterate on. https://arxiv.org/abs/1912.01603The TalkRL podcasts on this line of work are reasonable accessible and quite interesting.

      benbenben111Hacker News11d agoview on Hacker News ↗
    • Intuitively I wouldn't readjust how many steps they each do, but instead add another run afterwards, that get the same amount of steps as the previous, but now also with a concise description of what the previous attempts did and what they achieved, and ask it to improve. The amount of compute you have available, would dictate how many full…

      embedding-shapeHacker News11d agoview on Hacker News ↗
    all of them →
  2. 4

    'The Last AI Built by Humans' paper resurfaces on HN

    The same 'Last AI Built by Humans' paper is resubmitted to Hacker News by a different user hours later, indicating renewed attention to the topic.

  3. 3

    Researchers detail MetaRSI-v1 framework for AI self-improvement

    A paper describing MetaRSI-v1 is posted, proposing a scheduled composition of three operators (Data-RSI, Harness-RSI, Model-RSI) that let a model improve data use, its own scaffold, and its parameters without external supervision; it is validated only on standard coding and closed-form science benchmarks.

    “Recursive self-improvement (RSI) lets a system improve the model-building machinery from its own failures, so every later model inherits the gain.”
    — MetaRSI-v1 paper authors
  4. 2

    'Meta^N' recursive self-improvement paper surfaces on HN

    A separate arXiv paper, 'Meta$^N$: Recursive Self-Improvement Through Emergent Depth,' is posted to Hacker News the same evening.

    1. 1 outlet first by HN Frontpage, 11d ago · read ↗

    2. first by arXiv cs.AI, 13d ago

  5. 1

    Paper 'The Last AI Built by Humans' posted to arXiv

    A paper titled 'The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement' is submitted to Hacker News, linking to its arXiv abstract page.

    1. first by arXiv cs.AI, 12d ago

What people are saying 10 voices from 1 site · best of 13 · verbatim