scosman

16 ms·

1. 241.▲
   by scosman 1y ago

   Don't forget praising a completely closed competitors with zero sense of irony.
2. 242.▲
   by scosman 1y ago

   Pdoc is great. I tried mkdoc and others, but pdoc was so much easier. One command and you're done. Trivial to add to a Github workflow & Github pages.
3. 243.▲
   by scosman 1y ago

   Me as engineer: it's wild a big corporation dictates a quality bar and limit API access for 3rd party software. Me as customer: oh man I'm sure glad stuff is reviewed to some quality bar and the OS limits API access.
4. 244.▲
   by scosman 1y ago

   it would!
5. 245.▲
   by scosman 1y ago

   Few things: - Security/Speed: I leave "approve CLI commands" on in Cursor. This functions as a whitelist of known safe commands. It only needs to ask if running a non-standard command, 99% of the time it can use tools. It wil
6. 246.▲
   by scosman 1y ago

   this should be a MCP server the agent can use and optimize on I have a MCP server that wraps developer tool CLIs (linting, tests, etc), but this would need a textual report instead of HTML. https://github.com/scosman/ho
7. 247.▲
   by scosman 1y ago

   Thank you. Also a paying Kagi user because I like the idea that it’s worth it to pay for a good service. Ripping off journalists/newspapers content goes against that.
8. 248.▲
   by scosman 1y ago

   Cerebras has OpenAI compatible "Qwen Code" support. ~4000 tokens/s. Qwen code's 480B param model (MoE) that's quite good. Not quite sonnet good, but speed is amazing. https://www.cerebras.ai/blog&#x2
9. 249.▲
   by scosman 1y ago

   Interesting quirk on first use: "`temperature` and `top_p` cannot both be specified for this model. Please use only one."
10. 250.▲
    by scosman 1y ago

    Sure so 2.5.509
11. 251.▲
    by scosman 1y ago

    Models have changes worthy of semver style major changes. Tokenizer, tool support, tool format, JSON modes, etc. Pipelines absolutely must change when these change. This thread is more about the minor number: not incrementing it when making
12. 252.▲
    by scosman 1y ago

    This is great. I made a download button a while ago. The Apple and Windows logos scale down, look great, and are easily identifiable. Tux is great, but just doesn't scale down. Tried about 10 variants to get one that is recognizable, b
13. 253.▲
    by scosman 1y ago

    Agreed, and sorry about that. Maybe edit the incorrect comment about "I can see kiln has copied over some of our code" for clarity. I get it was probably honest mistake, but hard not to reply when people are claiming I copied some
14. 254.▲
    by scosman 1y ago

    I read the code. I also remember writing the code and that comment. As disclosed: some prompt strings were taken and modified, but none of the code was. The original strings are using a templating library that we don't support, so thei
15. 255.▲
    by scosman 1y ago

    I personally wrote Kiln's SDG code myself -- no code was copied from here or anywhere else. Not sure where that claim is coming from, but it's not accurate. I might have taken some of the prompts and modified them. I didn't r
16. 256.▲
    by scosman 1y ago

    If anyone's interested in synthetic data generation, we've built a fully interactive visual tool for SDG. It supports generating hierarchical topic trees like other tools, but we do two things others don't: First: fully inter
17. 257.▲
    by scosman 1y ago

    If the weights have changed via training (they have) it’s a new model. This isn’t “hardware optimizations”. It’s additional training/new-weights.
18. 258.▲
    by scosman 1y ago

    Ugh. If the model name includes sem_ver version number, increment the version number when making a new release! Anthropic learned this lesson. Google, Deepseek, Kimi, OpenAI and others keep repeating it. This feels like Gemini_2.5_final_FIN
19. 259.▲
    by scosman 1y ago

    Great ad for deno. I hit a similar one the other day from pydantic. They make a MCP server for running sandboxed python code and the they did that… Python to WASM, and wasm running in deno.
20. 260.▲
    by scosman 1y ago

    Best news I've heard all day.
21. 261.▲
    by scosman 1y ago

    Unless this is actually made by the SWE Bench team, and I see no evidence it is, this name is incredibly poor form. Just adding "Pro" to someone else's name not only is infringing on their mark, but implying yours is superior
22. 262.▲
    by scosman 1y ago

    Spent 15 minutes the other day testing a patch I received that claimed to fix a bug (Linux UI bug, not my forte). The “fix” was setting completely fictitious properties. Someone has plugged the GitHub issue into ChatGPT, spat out an unteste
23. 263.▲
    by scosman 1y ago

    That industrial design sure looks familiar...
24. 264.▲
    by scosman 1y ago

    Just report them for copyright infringement. Apple will take the blatant cases down. The copy-cats that don't infringe are just a side effect of success, so... congrats? Regarding Reddit, it's a great place to find users but also
25. 265.▲
    by scosman 1y ago

    > Apple is aware of a report that this issue may have been exploited in an extremely sophisticated attack against specific targeted individuals.
26. 266.▲
    by scosman 1y ago

    I'm worried about your cloud bill... but good stability while being #1 on HN
27. 267.▲
    by scosman 1y ago

    Interesting. I’m in safari.
28. 268.▲
    by scosman 1y ago

    Usability and perf experience for me: - I'm getting about 5 FPS scrolling on a M4 Pro - Moving a "window" around takes 29% of my CPU, and renders at about 2 fps - I'm losing about 40% of my screen height for reading (14&
29. 269.▲
    by scosman 1y ago

    Exactly. And most folks are using a framework like llama.cpp which does control where it’s run.
30. 270.▲
    by scosman 1y ago

    Anyone interested in this space should look at https://turbopuffer.com - I think they were first to market with S3 backed vector storage, and a good memory cache in front of it.