tough

26 ms·

1. 481.▲
   by tough 1y ago

   yeah idk but something like curation is done by commitee (to try and maintain a minimum of quality overall) but as an author is super easy to appply/get editors/feedback/publish (that’d be the self-serve part i guess) tbh har
2. 482.▲
   by tough 1y ago

   Ive seen claude 4 do this too when its context has lots of teats already and tool calling imho the main issue is an llm no has real sense of what’s a real tool call vs just a log of it, the text logs are virtually identical, ao the Llm star
3. 483.▲
   by tough 1y ago

   oh interesting wasnt really aware so you could use beam to orchestrate go or rust services communicating over IPC? Nice
4. 484.▲
   by tough 1y ago

   is that a fair comparison? i dont love k8s but you can deploy anything to it, not just erlang or elixir
5. 485.▲
   by tough 1y ago

   someone should build a cooperative/self serve after stripe press books
6. 486.▲
   by tough 1y ago

   local open source llms can run on your HW its nit really choosing one ot the other imho they supplement each other
7. 487.▲
   by tough 1y ago

   There's always XTool https://github.com/xtool-org/xtool
8. 488.▲
   by tough 1y ago

   I settled using latex with tectonic, you could always leverage playwright or similar for easy html -> print to pdf without any weird libs? (not great startup time, but you can batch many ops in one session) # justfile ── put in repo root
9. 489.▲
   by tough 1y ago

   You can use openai's codex, or any of the claude code clone's available on github. At the end of the day all do the same. a cli tool that can call tools and use them as needed. https://github.com/openai/codex
10. 490.▲
    by tough 1y ago

    What's the problem with making your own layouts? just the syntax?
11. 491.▲
    by tough 1y ago

    llms are great at handling syntax for you so you can focus on your work I never wrote latex before, but writing a simple PDF / scholarly article from code is pretty easy with current tools if you're a dev
12. 492.▲
    by tough 1y ago

    why not put claude code inside a docker environment you control? you might copy the code working dir in read mode only or whatever idk\
13. 493.▲
    by tough 1y ago

    It can do all this and more, you just have to nudge/instruct it to. It's also amazing at using the CLI (I've been a terminal nerd for years, and yet watching claude 4 pipe those unix commands to work around the fact it can&#x
14. 494.▲
    by tough 1y ago

    Thanks I never had learned non-rivalrous goods and its'a. great way to explain the IP != property you wouldn't steal a car, err i mean download, right!
15. 495.▲
    by tough 1y ago

    anyhow and thisiserror are my rust error lib defaults nowadays
16. 496.▲
    by tough 1y ago

    go has a really good ecosysstem between wrapcheck, gocritic, and golint
17. 497.▲
    by tough 1y ago

    Interesting approach! thanks for sharing
18. 498.▲
    by tough 1y ago

    Awesome, thanks! I do keep running on the issues, but the levers as you say make it harder to implement. As of right now, I could only replace the JSON tool calling on LLM's on something I fully control like vLLM, and the big labs prob
19. 499.▲
    by tough 1y ago

    Thanks yes! Im aware of structured outputs, llama.cpp has also great support with GBNF and several languages beyond json. I've been trying to create go/rust ones but its way harder than just json due to all the context/state
20. 500.▲
    by tough 1y ago

    In my experience with other regular models, once the context starts to fill up, quality starts to degrade. wouldn't getting batched at the end of a batch, have a similar -effect- on the results, where your prompt might recieve overall
21. 501.▲
    by tough 1y ago

    Malware is just a niche of adware anyways, all the ads are mostly vectors for malware to get into your computer. Use adblcok, Use firefox, be free
22. 502.▲
    by tough 1y ago

    Goboscript was on HN a few days ago https://news.ycombinator.com/item?id=44026799 for when they grow up =)
23. 503.▲
    by tough 1y ago

    Fake AI tools aren't AI Malware I wont even click the link and give them a visit
24. 504.▲
    by tough 1y ago

    I remember doing offline debit card payments 10y ago in a flight they would pass the card with one of those old engraving things lol
25. 505.▲
    by tough 1y ago

    how do you do that exactly?
26. 506.▲
    by tough 1y ago

    hi dan! really interesting post. do you think a new data serialization format built around easier generation/parseability and that also happened to be streamable because its line based like jsonld could be useful for some?
27. 507.▲
    by tough 1y ago

    that does make sense, sadly firecracker seems to be mostly relegated to linux for now so there's no good multi-arch story i'm aware of
28. 508.▲
    by tough 1y ago

    sorry i meant to ask simon directly if they require a non-docker solution im working on a wrapper that lets you swap runtimes and my first implementation is mostly a wrapper around docker containers planning to add firecracker next will exp
29. 509.▲
    by tough 1y ago

    I had luck using ALVM which users Apple Hypervisor framework while exploring linux micro-vm's in macos fwiw https://github.com/mathetake/alvm
30. 510.▲
    by tough 1y ago

    would you be OK with a -hardened- with default profiles docker containers one?