sevagh

46 ms·

1. 271.▲
   by sevagh 3y ago

   > LLM tools - Ollama, llama.cpp I was including these since they're LLM-related cli tools.
2. 272.▲
   by sevagh 3y ago

   >I'm puzzled that more people aren't loudly exploring this space (LLM+CLI) - it's really fun. 70% of the front page of Hackernews and Twitter for the past 9 months is about everybody and their mother's new LLM CLI. It
3. 273.▲
   GUI for a Vocal Remover That Uses Deep Neural Networks(github.com)

   1 points by sevagh 3y ago|0 comments
4. 274.▲
   by sevagh 3y ago

   I wrote pq [1] (protobuf parser cli) at a company where I was told to "just use the tool another engineer wrote" which was in C++, in a really uncompileable/abandoned/unusable state I wrote goat [2] (EBS disk attacher) a
5. 275.▲
   by sevagh 3y ago

   Example: Kubernetes, convincing 60% of the industry that they don't know how to run production workloads anymore.
6. 276.▲
   by sevagh 3y ago

   Not sure if it's a conspiracy theory or a known thing but I read it once and it stuck in my mind that Google's projects are a form of distraction and friction to prevent people catching up with them (e.g. HTTP3).
7. 277.▲
   Frugally-deep: Header-only library for using Keras (TensorFlow) models in C++(github.com)

   3 points by sevagh 3y ago|0 comments
8. 278.▲
   by sevagh 3y ago

   Hi all, I showed freemusicdemixer.com before with great reception. I finally got the state-of-the-art model, Demucs v4 hybrid transformer, running in the WASM environment! Happy demixing!
9. 279.▲
   by sevagh 3y ago

   Hi HN, I finally got the excellent Demucs v4 Hybrid Transformer model for music source separation running client-side with WASM, and added it to my website: https://freemusicdemixer.com This should be a huge step up in the separ
10. 280.▲
    by sevagh 3y ago

    >Larian, Grinding Gear Games, Remedy, Valve, and CD Project red This is a rollercoaster of a list. Valve's $7.7bil versus a company who's only product is Path of Exile?
11. 281.▲
    by sevagh 3y ago

    Around here (Montreal), cabs were miserable scammers and scumsuckers that would drive you to ATM machines in the middle of the night to coerce you to withdraw cash to pay them. I'm glad Uber fucked them to shreds. Now, of course, they&
12. 282.▲
    by sevagh 3y ago

    >I mean, yeah. How many engineers do you need? 5-6 for the backend, 3-4 DevOps, maybe some more for the various frontends... say, 20? So Spotify should bring you in as the CTO, right? How can you make such a confident claim that you know
13. 283.▲
    by sevagh 3y ago

    I still don't understand why the Quora UI shows you unrelated answers, and it's always a weird topic. "Question: Why did PyTorch beat Tensorflow in popularity" "Answer (to a different question): I lost my virginity
14. 284.▲
    by sevagh 3y ago

    No, and I don't think any CI platform deals with it in a great way. I mean that once you account for the fact that your dev team may not have access to all of the secrets, the single bash script/Makefile stops being able to shield
15. 285.▲
    by sevagh 3y ago

    >- Put all build/publish/test logic in Makefiles or scripts in the repo. This means that devs can run it locally as well. The only interaction between a test runner and the codebase should be running `make <target>`. This
16. 286.▲
    by sevagh 3y ago

    >It's not simple. Neither is using virtualenvs for Python packages with native extensions.
17. 287.▲
    by sevagh 3y ago

    Presumably it also can fill the role of conda/mamba i.e. also managing C/C++ libraries in the same way in the nix environment, isolated from the system libraries?
18. 288.▲
    by sevagh 3y ago

    Firejail is cool: https://github.com/netblue30/firejail Linux namespaces/cgroups but lighter than Docker. I use it when I want to limit the memory of a Python script: ``` maxmem="56" #GB firejail --nopro
19. 289.▲
    by sevagh 3y ago

    Yes, this really irritates me. Me: <prompt 1: modify this function> AI assistant (either ChatGPT or Copilot-X): <attempt 1> Me: <feedback> AI assistant: <attempt 2>, fixed with feedback, but deleted something crucial
20. 290.▲
    by sevagh 3y ago

    >I've heard people joke that AI is just matrix multiplication Not sure where the joke is, today's deep learning based AI models are literally matrix multiplications with learned weights.
21. 291.▲
    by sevagh 3y ago

    Yeah, agreed. I guess "bring the heat" implies combat but standing one's ground calmly (or, refusing to accept the person's heat) is better.
22. 292.▲
    by sevagh 3y ago

    Prometheus/Grafana doesn't have batteries included. Once you set up exporters and alerts and graphs, then you get something useful.
23. 293.▲
    by sevagh 3y ago

    If it's anything like companies I've worked at, you stick them in an unimportant fiefdom and give them raises, promos, and refreshers for decades.
24. 294.▲
    by sevagh 3y ago

    If they bring the heat, you bring it right back. If you let people bully you once it won't ever stop.
25. 295.▲
    by sevagh 3y ago

    I thought it was chasing cosmetics, size, and quantity (e.g. giant bland tomatos, overfed hormone chicken) that led to nutrient deficiency.
26. 296.▲
    by sevagh 3y ago

    This just got added to some lists of interview questions.
27. 297.▲
    by sevagh 3y ago

    This doesn't rule out that there were enough good or "better" applicants that crowded yours out.
28. 298.▲
    by sevagh 3y ago

    I believe another factor is that sometimes the model responds better to your prompt than other times. This way you get two dice rolls of your prompt hitting "the good path."
29. 299.▲
    by sevagh 3y ago

    Total Compensation
30. 300.▲
    by sevagh 3y ago

    I frequently share responses between ChatGPT (paid version with GPT4) and Copilot-X to break an impasse when trying to generate or fix a tricky piece of code.