hu3

12 ms·

1. 331.▲
   by hu3 10mo ago

   Check context size. LLMs become increasingly error-prone as their memory is fills up. Just like humans. In VSCode Copilot you can keep track of how many tokens the LLM is dealing with in realtime with "Chat Debug". When it reaches
2. 332.▲
   by hu3 10mo ago

   https://archive.ph/vp6rp
3. 333.▲
   by hu3 10mo ago

   The value depreciation of the hardware alone is going to be significant. Probably enough to pay for 3x ~$20 subscriptions to OpenAI, Anthropic and Gemini. Also, if you use the same mac to work, you can't reserve all 128GB for LLMs. Not
4. 334.▲
   by hu3 10mo ago

   Open weight models are neat. But for SOTA performance you need specialized hardware. Even for Open Weight models. 40k in consumer hardware is never going to compete with 40k of AI specialized GPUs/servers. Your link starts with: > &
5. 335.▲
   by hu3 10mo ago

   I love Exceptions as control flow! Thanks for the suggestion. I too use specialized exceptions. Some have friendly messages that can be displayed to the user, like "Please fill the password". But critical exceptions display a gene
6. 336.▲
   by hu3 10mo ago

   I like this. But doesn't it make sense to categorize en Exception thrown as an erro somehow? And a new user registration as an email info? Perhaps use tags then?
7. 337.▲
   by hu3 10mo ago

   Nice. I guess you write logs on the "final" block of a global try/catch/final? Something like: try { // handle request code } catch (...) { // add exceptions to log } final { &#x2F
8. 338.▲
   by hu3 10mo ago

   Thanks. I think I could use skills as "instructions I might need but I don't want to clutter AGENTS.md with them".
9. 339.▲
   by hu3 10mo ago

   I made a small mcp script for database with 3 tools: - listTables - getTableSchema - executeQuery (blocks destructive queries like anything containing DROP, DELETE, etc..) I wouldn't trust a textual instructions to prevent LLMs from dr
10. 340.▲
    by hu3 10mo ago

    utf-8
11. 341.▲
    by hu3 10mo ago

    iOS from 2 versions prior don't get latest Safari? I can't check because my wife's iPhone is, regrettably according to her, "updated to the latest glAss version".
12. 342.▲
    by hu3 10mo ago

    How can you defend Safari rendering broken sites for long periods due to lack of frequent updates as a good thing? The ever current adage of distortion field applies here. Just like Safari not having webgpu was touted as a feature and now t
13. 343.▲
    by hu3 10mo ago

    At least MCPs can be unit tested. With Skills however, you just selectively append more text to prompt and pray.
14. 344.▲
    by hu3 10mo ago

    Perhaps you could help me. I'm having a hard time figuring out how could I leverage skills in a medium size web application project. It's python, PostgreSQL, Django. Thanks in advance. I wonder if skills are more useful for non cr
15. 345.▲
    by hu3 10mo ago

    Many languages use this idiom. Some popular ones even. So while it's good that Rust joined them, it's hardly a differentiator.
16. 346.▲
    by hu3 10mo ago

    and it was 40k
17. 347.▲
    by hu3 10mo ago

    Fellow diagnosed ADHD here. And I know every ADHD is different and people are different. What helps me is: - Prefer faster models like VSCode's Copilot Raptor Mini which, despite the name, is like 80% capable of what Sonnet 4.5 is. And
18. 348.▲
    by hu3 10mo ago

    I worry about that too. But at this point, it's like refusing to use vehicles to travel long distances in fear of becoming physicaly unfit. We go to the gym.
19. 349.▲
    by hu3 10mo ago

    In a world where people scratch sports cars painting out of pure envy. It's hard to criticise someone trying to stay under the radar of the crazy.
20. 350.▲
    by hu3 10mo ago

    I think it's because the ground in frontend dev is shaky. Browsers keep changing along with people's expectation of what a website or app should be capable of. Meanwhile on the backend land, the same MVC framework from 2 decades a
21. 351.▲
    by hu3 10mo ago

    What baffles me is when people defend this broken wallet garden to be used by their grandmas. Surely they won't be able to tell which one of the 20 ChatGPT apps is the correct one in search results.
22. 352.▲
    by hu3 10mo ago

    > only a small fraction will be needed in VRAM at any given time I don't think that's true. At least not without heavy performance loss in which case "just be memory mapped" is doing a lot of work here. By that logic
23. 353.▲
    by hu3 10mo ago

    React only renders. You need a lot more than that for an app. And if you glue packages together you're basically reinventing frameworks. This can be a good thing or a bad thing. Regardless of your choice between glueing libs or a frame
24. 354.▲
    by hu3 10mo ago

    Sure. If by ignore you mean flaunt about Apple Intelligence only to fail miserably on the expectation they themselves generated.
25. 355.▲
    by hu3 10mo ago

    Bun can generate single file executables even from TypeScript: bun build ./cli.ts --compile --outfile mycli So the value proposition here is debatable. https://bun.com/docs/bundler/executables Also sta
26. 356.▲
    by hu3 10mo ago

    Mozilla has pressure from their sugar daddy, Google, to weaken ad-blockers.
27. 357.▲
    by hu3 10mo ago

    What kind of work do you do that CANT BE divided enough into tasks that can be accomplished mostly through prompting?
28. 358.▲
    by hu3 10mo ago

    No language is going to do that for you. And I don't think Go promissed otherwise. Perhaps it's about managing expectations.
29. 359.▲
    by hu3 10mo ago

    why isn't that the default? why ship broken implementation prematurely?
30. 360.▲
    by hu3 10mo ago

    You expect Go to magically make systemd journaling exist in macOS? I can't even begin to comprehend the thought process here.