all2

12 ms·

1. 211.▲
   by all2 10mo ago

   Why? Why kowtow to people who don't care about your wellbeing or long term success?
2. 212.▲
   by all2 10mo ago

   There are ways around this. You can push the success rate close to 100% if you use chain of thought and a quorum selection. It isn't great, and it slows response times, but if 85% isn't good enough, you just need to flip the coin
3. 213.▲
   by all2 10mo ago

   I'd be curious to hear more about this 'small model calls out to larger model' thing. Are there demos out there showing this in use?
4. 214.▲
   by all2 10mo ago

   That sounds like an opportunity for some inspection; coverage, linting (type checking??), and a by-hand spot check to assess the quality of testing. You might also inspect the QA process (ride-along with folks from QA).
5. 215.▲
   by all2 10mo ago

   I would consider feature complete with robust testing to be a great proxy for code quality. Specifically, that if a chunk of code is feature complete and well tested and now changing slowly, it means -- as far as I can tell -- that the abst
6. 216.▲
   by all2 10mo ago

   I did not. Thanks for the heads up!
7. 217.▲
   by all2 10mo ago

   This has been my major concern, so much do that I'm going to be launching a tool to handle this specific task: agent conception and testing. There is so little visibility in the tools I've used that debug is just a game of whackam
8. 218.▲
   by all2 10mo ago

   I'd rather just get a cat. :D
9. 219.▲
   by all2 10mo ago

   Fascinating. That UI looks like Vercel. What node/graph library are you using? It looks like it would be really useful for some projects I'm working on.
10. 220.▲
    by all2 10mo ago

    The teacher example was from one of those pop-psych books on being more efficient with one's time. I can't remember the title off the top of my head. Another example in the book applied the author's model of thinking to a pla
11. 221.▲
    by all2 10mo ago

    Thoughtfulness is sometimes increased by touch time. I've seen various examples of this over time; teachers who must collate and calculate grades manually showed improved outcomes for their students, test techs who handle hardware beco
12. 222.▲
    by all2 10mo ago

    I've found success using the following process: 1. get the LLM to generate a spec from the code (the spec is your new source code). Take your time here, make sure it is correct. Split out implementation details unless they are critical
13. 223.▲
    by all2 10mo ago

    I dunno what it is, but whenever I see it I feel as if I have been there before.
14. 224.▲
    by all2 10mo ago

    Hm. What's the difference between the premium and the standard recipes?
15. 225.▲
    by all2 10mo ago

    My email is in my profile. Feel free to reach out.
16. 226.▲
    by all2 10mo ago

    The people developing standards are in the business of developing standards. It makes sense to want to make money on the thing you work on.
17. 227.▲
    by all2 10mo ago

    These are standard business practices. They own IP. People want to use that IP. They say "pay us X to use our IP". People throw a tantrum because money. Instead, people want to capitalize on someone else's hard work for free.
18. 228.▲
    by all2 10mo ago

    The tracking app would be pretty simple. A shim in the shell that tracks the first argument on the command line and increments a counter. Then usage stats are pushed to the service every 24 hours (or something like that). app usage / a
19. 229.▲
    by all2 11mo ago

    288GB RAM on board, and RISC V processors to enable the option for offloading inference from the host machine entirely. It sounds nice, but how much is it?
20. 230.▲
    by all2 11mo ago

    I take great pride in my role as a user of HN. This includes never reading the article and always positing very opinionated tasks without qualification.
21. 231.▲
    by all2 11mo ago

    This was one of my favorite movies growing up. It made me want to get into aerospace stuff. This and star wars.
22. 232.▲
    by all2 11mo ago

    Interestingly, I canceled my Claude subscription. I've paid through the first week of December, so it dries up on the 7th of December. As soon as I had canceled, Claude Code started performing substantially better. I gave it a design s
23. 233.▲
    by all2 11mo ago

    There's so many moving pieces in this mess. We'll normalize on some 'standard' eventually, but for now, it's hard, man.
24. 234.▲
    by all2 11mo ago

    Do not be deceived, bad company corrupts good morals.
25. 235.▲
    by all2 11mo ago

    I was hoping to point more towards "don't suppress a viewpoint, rather discuss it" and less toward semantics. I guess I should have illuminated that in my above comment.
26. 236.▲
    by all2 11mo ago

    > ethically all positions are not equally valid. Ethically, no; morally, yes.
27. 237.▲
    by all2 11mo ago

    I just use claude code for most things. I'll fall back to a web UI (Grok, Claude, or Gemini, depending on what service I've exhausted) if I need to.
28. 238.▲
    by all2 11mo ago

    Specifically thin sliced potatoes fried in industrial lubricants.
29. 239.▲
    by all2 11mo ago

    I have. I get into the same head space, either way.
30. 240.▲
    by all2 11mo ago

    These are intertwined, though, and rather tightly in some cases. Game dev is an excellent example of this.