smcleod

16 ms·

1. 301.▲
   by smcleod 1y ago

   Still no reduction in price for models capable of Agentic coding over the past year of releases. I'd take the capabilities of the old Sonnet 3.5v2 model if it was ¼ the price of current Sonnet for most situations. But instead of releas
2. 302.▲
   by smcleod 1y ago

   I think this is good, I'd like to see more side projects that don't directly relate to the core web browser shutdown in favour of browser performance, compliance and feature parity.
3. 303.▲
   by smcleod 1y ago

   Looks like it's a browserwalled to users of Chrome or Safari and promotes their use as an "upgrade", blocking all Firefox users. "Improve your experience" "Please use the latest version of a supported
4. 304.▲
   by smcleod 1y ago

   Perhaps they simply are someone who struggles with finding identity and value in change and adaptation.
5. 305.▲
   by smcleod 1y ago

   For me it decreases the barrier to try and test new thoughts, never have I felt more empowered to try out new avenues that in the past might have been too time consuming or expensive to dispose of.
6. 306.▲
   by smcleod 1y ago

   I very much agree with your sentiment here. I tried to encapsulate that to some degree when writing something (perhaps poorly?) recently actually - https://smcleod.net/2025/03/the-democratisation-paradox-what...
7. 307.▲
   by smcleod 1y ago

   Early on I wondered if things like this might happen. But for me what has actually happened is almost the opposite, I seem to be experiencing more of a "tree of thoughts" with the ability to now perform rapid experimentation down
8. 308.▲
   by smcleod 1y ago

   Yeah I get that, that's what we yse k/v cache quantisation for now which has a lower impact on PPL than this unless I'm missing something?
9. 309.▲
   by smcleod 1y ago

   +0.86% perplexity it's quite a bit at such a small context size though isn't it? How is it at more reasonable context sizes like 64-128k?
10. 310.▲
    by smcleod 1y ago

    GPT 4o and 4.1 are both pretty terrible for coding to be honest, try Sonnet 3.7 in Cline (VSCode extension). LLMs don't have up to date knowledge of packages by themselves that's a bit like buying a book and expecting it to have u
11. 311.▲
    by smcleod 1y ago

    Definitely recommend docling for this. https://docling-project.github.io/docling/
12. 312.▲
    by smcleod 1y ago

    I would not recommend basing your core business capabilities around an external chat app with no redundancy. Outages happen and always will.
13. 313.▲
    by smcleod 1y ago

    major incident acknowledged https://slack-status.com/2025-05/7b32241eb41a54aa
14. 314.▲
    by smcleod 1y ago

    Slack, Fastmail, Claude, Github - all seem to be having intermittent outages and are randomly logging people out.
15. 315.▲
    by smcleod 1y ago

    I'm not sure I understand why an open source maps app needs shareholders in the first place?
16. 316.▲
    by smcleod 1y ago

    Way back when, in physical land - I used STONITH for that! https://smcleod.net/2015/07/delayed-serial-stonith/
17. 317.▲
    by smcleod 1y ago

    I had a go at writing a bit of a sampling guide for Ollama/llama.cpp as well recently, open to any feedback / corrections - https://smcleod.net/2025/04/comprehensive-guide-to-llm-sampl...
18. 318.▲
    by smcleod 1y ago

    It's not immediately clear to me if this is a tool for ages (e.g. a MCP server), a browser engine (e.g. browserless) or some sort of OpenAI compatible LLM proxy that injects a web browser tool? It appears to expose itself via an openAI
19. 319.▲
    by smcleod 1y ago

    All these streaming services need better down mixing from surround sound to stereo, so often people can't hear the dialogue because it's been mixed for a centre speaker which becomes too low in the downmix to 2.0/2.1.
20. 320.▲
    by smcleod 1y ago

    You and I both!
21. 321.▲
    by smcleod 1y ago

    Everyone makes mistakes. It's important to make it easy to do the right thing.
22. 322.▲
    by smcleod 1y ago

    The EU has done lots of reach on road and car safety, there's lots of data out there - just perhaps not in the US as many American made cars have significantly lagged behind in terms of safety features.
23. 323.▲
    by smcleod 1y ago

    I don't understand what you're saying here - you think it's good to have your lights turned on - but you don't want them to automatically turn on?
24. 324.▲
    by smcleod 1y ago

    To be honest most of those accessories are actually incredibly cheap at manufacturing time and several have a direct impact on safety (e.g. ensuring people don't drive around with lights off). The cost usually comes as companies use th
25. 325.▲
    by smcleod 1y ago

    Also GLM 4 is pretty amazing - https://www.reddit.com/r/LocalLLaMA/comments/1k4w9p2/i_uploa...
26. 326.▲
    by smcleod 1y ago

    Hopefully Qwen 3 and maybe if we're lucky Qwen 3 Coder might be out this week too.
27. 327.▲
    by smcleod 1y ago

    That's a really old model now. Even the old Qwen 2.5 coder 32b model is better than DSv2
28. 328.▲
    by smcleod 2y ago

    Out of interest have you found it to work well with any locally hostable models (reliable tool calling, decent coding etc...)?
29. 329.▲
    by smcleod 2y ago

    The extension seems awfully heavy and it loads every time you run VSCode - not just when you need it.
30. 330.▲
    by smcleod 2y ago

    Neat! Is the source available somewhere?