brookst

26 ms·

1. 721.▲
   by brookst 5mo ago

   You seem more invested in hating RAM manufacturers than interested in the actual economics of the business. Look up Qiminda, ProMOS, Elpida. They invested in capacity during booms.
2. 722.▲
   by brookst 5mo ago

   Not the person you’re replying to, but RAM has historically been a boom-or-bust business, and companies that invest to meet demand during a boom cycle usually have that new capacity come online just in time for the bust. If it was just vari
3. 723.▲
   by brookst 5mo ago

   Same here. And I find that inviting research and dissent makes it even stronger. “I’m thinking we need to model prompt assembly as a graph, with versioning for graph configs. Please do some research on best practices in this area and see if
4. 724.▲
   by brookst 5mo ago

   Probably not that exactly, but there is a tradeoff between effectiveness of the prompt and cache hit rate. If putting the user’s datetime in the middle of the prompt scores higher on evals but worsens cache hits, versus at the end of the pr
5. 725.▲
   by brookst 5mo ago

   Cache is always there, it’s just that it only caches up to the point where an input token changes. So if the tools list is early in the prompt, changing it would limit cache for most of the prompt. If the tools list is the last thing, you c
6. 726.▲
   by brookst 5mo ago

   You pay, at 10% the price (in quota or dollars) for non-cached. See https://platform.claude.com/docs/en/about-claude/pricing
7. 727.▲
   by brookst 5mo ago

   Ah, understood, and thanks for the clarification!
8. 728.▲
   by brookst 5mo ago

   Claude code gets >98% KV cache hits. It’s not reprocessing unless you let the cache go cold (5 minutes, which is annoyingly short).
9. 729.▲
   by brookst 5mo ago

   Yep. I get $6k - $8k worth of tokens (at api rates) using the $200 max subscription.
10. 730.▲
    by brookst 5mo ago

    I get 98.6% cache hits on Claude code. Short of drastic arch changes it’s hard to imagine it getting much better.
11. 731.▲
    by brookst 5mo ago

    Are you saying Congress should be domain experts every area they allocate funds to? From maritime safety to preschool nutrition, congress should be expert enough in everything to specify all important details, and then the agencies (staffed
12. 732.▲
    by brookst 5mo ago

    How does that work with torrents?
13. 733.▲
    by brookst 5mo ago

    More advocacy I’d say, but interesting. How do pitches for charities, political campaigns, etc, change when the messaging is to agents?
14. 734.▲
    by brookst 5mo ago

    They really are at the forefront of legacy system support, sometimes starting it just months after a product launches.
15. 735.▲
    by brookst 5mo ago

    IMO there is a lot of daylight between “is not perfectly capable of simulating all situations and always used perfectly to the full capabilities of the system” and “doesn’t mean much”.
16. 736.▲
    by brookst 5mo ago

    I’m going to steal that
17. 737.▲
    by brookst 5mo ago

    100% agreed. I have yet to hear a convincing argument for why it is creative accretion when I leverage all of the music I’ve ever listened to in order to write an “original” song, but its base plagiarism when AI does similar. The only remot
18. 738.▲
    by brookst 5mo ago

    Technically yes, but my point was that I do not think Anthropic would be receptive to a pitch to fine tune Opus to be an edgelord.
19. 739.▲
    by brookst 5mo ago

    If you’re asking some some kind of abstract moral value sense, I have idea. If you’re asking whether project leads give more weight to a single, tangible, vocal stakeholder than they do to unknown numbers of anonymous and lightly-engaged st
20. 740.▲
    by brookst 5mo ago

    I too am outraged that a product would prioritize its biggest users.
21. 741.▲
    by brookst 5mo ago

    I don’t see where they charge for it.
22. 742.▲
    by brookst 5mo ago

    I don’t think we’ll ever entirely know what they mean.
23. 743.▲
    by brookst 5mo ago

    Can you elaborate on how you use the zero? I got all excited, bought one, and it’s in a drawer. I’m way deep into coding, CNC machining, making of all sorts… but I just never incorporated it. What am I missing? What do you use yours for?
24. 744.▲
    by brookst 5mo ago

    It reminds me of high school, ages ago, when a friend would go on and on about how Depeche Mode weren’t musicians and how nobody cares about electronic music. I’m a little nostalgic for the hours, cumulatively probably weeks, that I heard a
25. 745.▲
    by brookst 5mo ago

    My take is that it’s about better replicating products that use water. San Francisco bread for instance. So it’s not about the best water to drink.
26. 746.▲
    by brookst 5mo ago

    While I appreciate the condescension, tell me how Walmart is a monopoly?
27. 747.▲
    by brookst 5mo ago

    Grok’s whole thing is being irresponsible and edgy. Can’t s ee Anthropic going for that, can’t see Grok’s customer base accepting an AI that refuses to do nonconsensual porn.
28. 748.▲
    by brookst 5mo ago

    Well, brikym, exactly how hot is this hot coffee? If it’s within normal expectations for coffee it is likely that person’s fault. If it is 210 degrees F, it is likely McDonald’s fault.
29. 749.▲
    by brookst 5mo ago

    The “we’s” are different. Some of us will use AI to replace human relationships and our own decision making, others of us will use it to make amazing art and invent new things.
30. 750.▲
    by brookst 5mo ago

    Thank you for the Watch the Sound recco, looks great.