The Miller Principle (2007)

(puredanger.github.io)

45 points | by FelipeCortez 4 days ago

21 comments

  • donatj 19 minutes ago
    For fun, I recently rebuilt a little text adventure some friends and I had built in the early 2000s. Originally written in QBasic, I translated it line by line in Go, and set it up as a little SSH server.

    For posterity, I didn't want to change anything about the actual game itself but knew beforehand that the commands were difficult to figure out organically. To try and help modern players, I added an introductory heading when you start playing informing the player that this was a text adventure game and that "help" would give them a basic list of commands.

    Watching people attempt to play it in the logs, it became painfully obvious no one read the heading, at all. Almost no one ever typed "help". They'd just type tens of invalid commands, get frustrated and quit.

  • sdevonoes 1 hour ago
    I think this is more true now than ever. Before LLMs, when someone came up with an ADR/RFC/etc you had to read it because you had to approve it or reject it. People were putting effort and, yeah, you could use them in your next perf. review to gain extra points. You could easily distinguish well written docs from the crap (that also made the job of reviewing them easie)

    Nowadays everyone can generate a 20-page RFC/ADR and even though you can tell if they are LLM generated, you cannot easily reject them based on that factor only. So here we are spending hours reading something the author spent 5 min. to generate (and barely knows what’s about).

    Same goes for documentation, PRs, PRs comments…

    • manmal 2 minutes ago
      Those generated ADRs are pure crap, full of unnecessary hedges and superficial solutions that don’t survive scrutiny longer than 10 seconds. I do generate ADR skeleton drafts because I hate empty pages, but I need to add the substance or they are not helpful at all.

      What we are doing is probably not in training data, maybe that’s why.

    • ghgr 1 hour ago
      As a counterexample, thanks to LLMs many long-form articles that get posted with clickbaity (but devoid of content) headlines that I would have ignored otherwise now get "read" (albeit indirectly, with the prompt "Summarize the insights of the article $ARTICLE_URL in an academic, dry, technical and information-dense way")
      • eru 1 hour ago
        I notice that with YouTube videos.
    • jodrellblank 55 minutes ago
      Watching the Artemis II splashdown and following media event, I’m suspicious that a woman from TechTalk Media read out some LLM blurb instead of asking a question; I can’t prove it, but I can almost hear the em-dash in:

      "What you have done this week is remind the people of Earth that wonder is worth chasing. That curiosity is the most human thing we have. You didn't just test a spacecraft -- you tested mankind's potential...”

  • torben-friis 2 hours ago
    I wish this was the case. Then we wouldn't have a minority of us deeply frustrated :)

    'Thanks for the doc, let's set a meeting' (implied: so you can read the doc aloud to us ) is the bane of my existence.

  • sebastianconcpt 46 minutes ago
    This signals something that is happening somehow predictably due to the increasing abundance of code. It exponentially grows the surface offered for understanding (text as in comments, docs etc) and our attention bandwidth, well, is not exponentially growing, so...
  • comrade1234 2 hours ago
    Despite using an ai while programming I still have open Java doc and other api documents and find them very useful as the ai often gives code based on old apis instead of what I'm actually using. So I do read those documents.

    But also, I have a somewhat mentally ill (as in he takes medication for it) coworker that sends rambling extra-long emails, often all one paragraph. If I can't figure out what he's asking by reading the first couple and last couple of sentences I ask him to summarize it with bullet pouts and it actually works. Lol.

  • taffydavid 2 hours ago
    I read this entire post and all the comments this disproving the Miller principle
    • armchairhacker 1 hour ago

          This principle applies to the following:
      
          - User documentation
          - Specifications
          - Code comments
          - Any text on a user interface
          - Any email longer than one line
      
      Not blog posts or comments. Ironic
      • taffydavid 1 hour ago
        Damn, I guess I didn't read it closely enough
        • sebastianconcpt 1 hour ago
          Proved that read is not causation of understanding but mere correlation.

          So if the read of the Miller principle is interpreted as read+understanding (it should) an interesting deeper discussion can happen.

          It can be invoked with a way more dramatic "None understands anything"

  • coopykins 1 hour ago
    It's one of the main things I learned when working as tech support and I talked with users all day. Nobody reads anything.
    • funnybeam 1 hour ago
      I used to refer to the helpdesk as the reading desk - “Hello, you’re through to the IT Helpdesk, what can i read for you today?”
  • Borg3 30 minutes ago
    We are reaching society shown in "Johny Mnemonic" movie.. So much (useless) information around that people gets overloaded. I barely read anything these days on NH, too much (crap) information. I skim and only read stuff that is very close to my interest.

    I used to read a lot more in the past, not the case anymore..

  • hamdouni 2 hours ago
    Yeah, i'm also surprised people just read post title and jump to conclusions ...
  • Animats 2 hours ago
    The LLMs read everything.
    • krona 1 hour ago
      It doesn't mean they're paying attention.
    • formerly_proven 2 hours ago
      Only because they are architecturally unable to not read something.
  • stevage 1 hour ago
    Should probably be "The Miller Principle (2007)"
  • ekjhgkejhgk 1 hour ago
    Damn, this is thin content even for HN.

    Anyway, this is just projection. The Miller principle really should be "Miller doesn't read anything". I read plenty.

  • fmajid 1 hour ago
    Write-only memory
  • realaleris149 2 hours ago
    The agents will read them
  • smitty1e 2 hours ago
    I have found much value in reading the python and sqlite documentation. The Arch wiki is another reliable source.

    Good documentation is hard.

    • simultsop 1 hour ago
      I don't know. Under pressure and stress all docs are ugly.
    • Akcium 2 hours ago
      I would love to answer your comment but I didn't read it :P
  • spiderfarmer 2 hours ago
    The Laravel documentation is GREAT when you're getting started. Every chapter starts by answering the very question you might ask yourself if you're going through it top to bottom.

    I'm a one-man-band so if I write code comments, I write them for future me because up to this point he has been very grateful. Creating API documentation is also easy if you can generate it based on the comments in your code.

    Maybe rename it the Filler principle. Nobody reads mindless comments that are 'filler'.

  • makach 2 hours ago
    ..and emails
    • stevage 1 hour ago
      > Any email longer than one line

      it's in there

    • sarreph 1 hour ago
      The irony.
  • timrobinson33 1 hour ago
    tl;dr