Show HN: The Riemann Proof Handbook

(github.com)

82 points | by tamnd 16 hours ago

18 comments

  • brandonpelfrey 11 hours ago
    This feels like it has nowhere near enough massaging of the writing style, it’s as if they just told Claude to write a book and used defaults. You absolutely must tell models to speak in certain ways, use guidance to reduce Claudisms, and run adversarial review with other separate models that take a viewpoint of a uniformed reader. That is the only way to start getting semi-reasonable writing for something like this.
    • sam345 7 hours ago
      Can you give some tips or links etc as to how to get Claude to stop writing like Claude. I have tried and it's difficult to get rid of the Claudisms consistently. really would be nice to have a simple prompt to give it.
    • thatsabadlook 10 hours ago
      So many people who are heavy users of llms do not care at all about the end result of their usage. They don't even read it even if something to be read is the deliverable.

      I can't tell if they think no one notices this. If they lack taste. Or if they just don't care.

      • stingraycharles 10 hours ago
        I actually feel like it’s the people that don’t use LLMs that much that do this, as the people that do use it all day surely must have a nearly visceral reaction when reading this?

        And heavy users of LLMs surely know exactly how to get it to write decently?

        • layer8 9 hours ago
          Some heavy users just get used to the stylistic abominations and stop perceiving it as something wrong.
          • hyperhello 8 hours ago
            Even though I agree with you, I also realize that “explain it to me like I’m on social media, not an obsessed nerd” is not going to happen when we’re talking about the Riemann proof. The subject is unsuited to the tone.
    • ctrlkctrls 9 hours ago
      I don't get this. It's all the time in HN and it's pathetic - the tone doesn't matter, can't you get over the tone and writing style and focus on the content instead?

      I swear we could discover proof of alien life and all HN would talk about would be the tone of the press release.

      • danlitt 7 hours ago
        What people call "tone" is the LLMs failing to actually communicate information. The explanations generated by LLMs by default do not actually explain, even if the content is true. That's why people get so angry about it.
      • Icathian 9 hours ago
        > I don't get this

        It sounds like you don't want to get this. If you do, I'm sure plenty of people would be happy to explain their positions on it.

      • skipkey 8 hours ago
        Honestly I think that some people just exist to complain on the internet. Lord knows I have been guilty of it myself in the past. This complaint is particularly silly though, because a) it’s not that hard to read, b) there is nothing stopping them from dropping it into Claude and saying “make it sound less like you”, and c) it’s actually a useful signal. If you know something is llm output you know there are probably small hallucinations in it and to absolutely double check anything from it before relying on it.

        Me personally I would be more likely to use a prompt like “try to make it understandable to a second year math student at university”.

      • ozgung 7 hours ago
        Last week there was no proof of the GRH. Two days ego the world didn’t know about the proof of the GRF. Yesterday everyone learned about it but no human was able to read the proof. Today someone magically created a full book on the proof and distributed for free. For me it’s impossible to read and understand it only because it’s very advanced math. But people’s first reaction to this is what, lack of polish in the tone?
  • Revanche1367 9 hours ago
    Mathematical proofs potentially becoming commodities in my lifetime is rather bewildering, it feels more like opportunities lost rather than gained for the common man though. Whoever has the most compute is likely to have the best or largest stock of proofs.

    Corporations and private companies are ahead right now, but soon enough nation states will likely resolve to run their own agent swarms, some are probably doing so already. Why wouldn’t they, I think this will likely trigger a real AI arms race, high caliber open models are not likely to remain accessible for much longer.

    • ahmedfromtunis 6 hours ago
      How is this different from the status quo? Don't these companies and governments already have the advantage of being able to hire the best minds and use them to their advantage?
      • Revanche1367 2 hours ago
        There was no mostly automated way to produce general proofs in mathematics before this except in theory. Automated theorem provers still needed a human to understand the proof and operate the program at least as far as getting it ready to start the analysis. Programs being proofs in the functional tradition is arguably the same thing from a philosophical perspective at least but that methodology alone hadn’t achieved the same level of automation until now with LLMs, most decisively shown by the lean proof for GRH I think. Even if not categorically different from the status quo, acceleration and pervasiveness does matter from a practical human standpoint.
  • Kuyawa 7 hours ago
    Computers helped us think faster. The internet helped us gather knowledge faster. AI helped us create faster. The information output is overwhelming. One big/little thing can be replicated and spread all over the world for everybody to improve on it, to modify, update, use, repurpose, share. And that is the vertiginous speed we are sensing right now, we are not used to this velocity of information flow

    FOMO is at is peak, you try to check all news not to miss the latest, this year alone a million things have happened, we are starting to live a compressed life, like a zip file of facts unzipped at once day after day

    • throw0101d 6 hours ago
      > Computers helped us think faster.

      Steve Jobs in 1990 in The Machine That Changed The World, "The Paperback Computer" (PBS-WGBH/BBC):

      > I remember reading an article when I was about twelve-years-old. I think it might have been Scientific American where they measured the efficiency of locomotion of all these species on planet earth. How many kilocalories did they expend to get from point A to point B? And the condor came in at the top of the list: surpassed everything else. And humans came in about a third of the way down the list, which was not such a great showing for the crown of creation. But somebody there had the imagination to test the efficiency of a human riding a bicycle. A human riding a bicycle blew away the condor all the way off the top of the list. And it made a really big impression on me that we humans are tool builders. And that we can fashion tools that amplify these inherent abilities that we have to spectacular magnitudes. And so for me, a computer has always been a bicycle of the mind.

      * https://americanarchive.org/catalog/cpb-aacip-15-qj77s7j327

      * https://en.wikipedia.org/wiki/The_Machine_That_Changed_the_W...

    • bongofizzbuzz 5 hours ago
      [flagged]
  • tito 15 hours ago
    I built a tool so anyone can run the Riemann proof on their spare CPU cycles.

    And there's a global leaderboard too (6 people running it now). Graphics inspired by SETI@home

    https://riemann-at-home.github.io/riemann-at-home/

    I’m here for all the tools to understand this Riemann Hypothesis breakthrough!

    • LiamPowell 11 hours ago
      I don't really understand the point of this, you don't need a distributed system, it only takes a few hours on any desktop PC.
      • tito 9 hours ago
        “Yep it compiles” was significant enough a statement that the Lean and Mathlib communities, endorsed by the Lean FRO and the Mathlib Initiative, issued a joint statement on it (Riemann@home validates items 2 and 4).

        https://news.ycombinator.com/item?id=50589618

        • LiamPowell 8 hours ago
          I understand what Lean does, I don't understand why you need a whole distributed system for something anyone can do themselves.
          • tito 7 hours ago
            Why not? What I like about it is it’s fun and sparks questions about the nature of trust and truth. Yes, so and so mathematicians said this runs in Lean. And I could run it on my home computer. So why not do so. And layer on a leaderboard while I’m at it so my friends can join in.
    • jychang 10 hours ago
      This took like 12 hours to build on my MacBook. What’s the point of this? It takes longer to build a Linux kernel on older machines. There’s no Linux kernel build leaderboard.
  • henearkr 14 hours ago
    I thought that the Lean project for the RH had been built and checked already several times with different kernels?
    • layer8 9 hours ago
      You thought correctly, and the Readme is wrong.
    • QuesnayJr 12 hours ago
      They have.
  • reader9274 15 hours ago
    I don't think any one person would have been able to link so many areas of math to come up with this proof in the ai-form. Masullo's proof is definitely more digestible and human-plausible.
    • rustyhancock 13 hours ago
      Something similar was said about Andrew Wiles FLT proof - i.e. that it was stunning a single person that traversed such vast mathematical territory.

      Over the past 30 years (apparently I can't testify to this personally) it's much slicker but for the most part it is still covers mostly the same range of vast mathematics.

      I do wonder if eventually someone will simplify it somewhat.

      Apparently, if the ABC conjecture is true FLT becomes a one pager proof.

      But the only "proof" we have for ABC is basically illegable with months of work and those who have taken the effort to try do not believe it to be correct.

      To attack this disagreement apparently the majority of the current effort is on using AI to translate it to familiar mathematics arguments and lean to verify.

      It feels romantic to say maybe the proof is correct but we're stuck between complex maths arguments people understand or alien seeming maths only one guy seems to understand.

      https://www.newscientist.com/article/2522687-the-secret-proj...

      • throwaway81523 12 hours ago
        I saw something about that pretty recently. The formalization effort hit a snag in the exact same place that Scholze and Stix did when they tried to review the ABC paper by hand. Unless a fix is forthcoming, things are as they were.
      • jychang 10 hours ago
        You’re out of date on the ABC conjecture.

        Multiple groups using AI have tried to formalize Mochizuki’s proof, and they all got stuck in the same place: Corollary 3.12, which is the same place the human mathematicians pointed out the flaw years ago.

        All evidence now points to that Corollary 3.12 is flawed, and Mochizuki’s proof would not work.

      • geordiew 13 hours ago
        < apparently...

        I believe this is true for asymptotic FLT (ie FLT for large enough n) and Mordell but not the full FLT.

    • eru 14 hours ago
      Alas, it's also incomplete, and perhaps even 'wrong'.
      • varjag 13 hours ago
        There's no source for that other than few randos' Claude sessions.
        • eru 12 hours ago
          The 'incomplete' is from the author himself.
          • varjag 10 hours ago
            Obviously. The "wrong" part though isn't.
  • irnbrudrinker 13 hours ago
    Has anyone scanned through and found any unknown insights Claude might have used? Or is the whole thing just stuff we could already do but was past the scale of a human's ability to do so?
    • QuesnayJr 12 hours ago
      From what I can tell, yes, there are new insights. The proof goes by a route that nobody had seriously considered. It's a fundamentally new proof technique, and may open up new frontiers. At the same time, it relies on a lot of existing math from the last 100 years, so it's not like humans accomplished nothing of relevance in that time.
  • jaen 8 hours ago
    Just a sea of Claudese slop, it's a rather grating and slow read. An insult to the reader, really.

    It's impossible to make Claude write fluent, coherent and correct English, even with instructions. (in a long-context setting – eg. "thinking" seems to exacerbate this)

    (with a massive tome of guidelines, Gemini Neon + GPT 6.1 Sol/Astra might approach something that doesn't make one want to vomit immediately.)

    • halJordan 1 hour ago
      The early ai period we're in is like the 50s for the tech we have today. You shitting on a rough first gen doesn't make you cool. If this were the 50s you'd be shitting the equivalent of your neighbor buying a fridge or a chambers oven. It doesn't make you cool
  • RiccardoSottini 8 hours ago
    Would it be a good idea to make a framework out of the runbook system you made, so you can then build other proof handbooks in the future?
  • Topology1 4 hours ago
    Your status page was outdated when you posted it, and still needs updating
  • loglog 9 hours ago
    It is easy to believe that your agent generated something useful. In fact, it did not. There are collaborative projects simplifying the proof right now, so whatever slop this might be is already outdated.
  • reader9274 15 hours ago
    I got the first 2 definitions. I'll take a break now.
  • eru 14 hours ago
    > Nobody has built the Lean project in full yet

    Huh? I thought multiple people did?

    • jychang 10 hours ago
      Yeah, I personally compiled the Lean source myself. It takes 5 mins to ask Codex to start the build, and it builds in a few hours.

      Not exactly impossible to do at home.

  • aniai 10 hours ago
    Wait hasn't it been done multiple time over?
  • bananaflag 14 hours ago
    > in PDF and EPUB

    Okay, so there should be prominent links to the PDF and EPUB.

    • alex-moon 13 hours ago
      At the very top of the README.md it makes it clear that the PDFs and ePUBs can be downloaded from the release artifacts, which view is linked directly. My humble opinion is that this is the most correct or at least standard way to do this on GitHub.
      • Timwi 11 hours ago
        I agree. I had no difficulty finding and opening one of the PDFs just for curiosity. (Of course I understood absolutely nothing in it)
  • 1970-01-01 7 hours ago
    It's not a hypothesis anymore. It's vetted logic. Riemann Theorem. No idea why Wikipedia is so slow in updating and still calling it the Riemann Hypothesis.

    https://simple.wikipedia.org/wiki/Theorem

  • Luke2642 12 hours ago
    I'll wait for the human rewrite.

    Rather than making my eyes bleed with claudisms, it needs completely rewriting. For example, the unitary hypotheses node page intro paragraph is:

    The Lean code ends the trunk with one existence statement. A record of forty-six fields, called HypsAlt, has an element. This page says what that record is, what the words "definite side" and "closed bound" mean in the code, and what the three large branches prove.

    But wouldn't an intro be so much nicer with a little narrative leading:

    This final node of the main proof shows that a 46-field record called HypsAlt has at least one element. The Lean theorem takes seven inputs, including sixty literature hypotheses on unitary groups and a "closed bound" on families of arithmetic data. Three main branches use these inputs to prove the "definite side" condition and two estimates, then build the HypsAlt element.

    • ninjalanternshk 10 hours ago
      What part of “the load bearing branches name their seams” is unclear to you?
      • r_lee 9 hours ago
        Now that's the smoking gun, not the seam.
        • fb03 2 hours ago
          that's the gate
      • Tanjreeve 9 hours ago
        Clearly anyone not finding that clear is part of some cabal of mathematicians who are just coping with an unstoppable force while also somehow holding back development of the technology
  • tyiz 12 hours ago
    [dead]