verdverm

19 ms·

1. 541.▲
   by verdverm 2mo ago

   Little Qwen nearly reaching the top 10 is certainly a supporting signal
2. 542.▲
   by verdverm 2mo ago

   https://archive.is/XwfoV
3. 543.▲
   by verdverm 2mo ago

   Perhaps the media is overlooking American Amnesia? or over indexing on what gets clicks, like they always do? I definitely tend to think that investors are much more sophisticated and nuanced in their analysis than the lost news media is
4. 544.▲
   by verdverm 2mo ago

   take a look at ADK, they have features for exactly this, mixing determ with agentic in a dag
5. 545.▲
   by verdverm 2mo ago

   this is why I use open weight models, LLMs are just not as good as people claim (though still very useful), but I'm spending way less Ant/Opus is all about that branding and 3-6 month lead, but it's not that much better than
6. 546.▲
   by verdverm 2mo ago

   I'm particularly fond of the concepts within git-appraise, gits back to the federated foundations https://github.com/google/git-appraise (unmaintained)
7. 547.▲
   by verdverm 2mo ago

   I would support time based quotas or limits per tier (free, solo, team, etc) But I'm not going to pay per commit over my monthly plan
8. 548.▲
   by verdverm 2mo ago

   Totally agree and this is actually on my short list. I use opencode, so one will need a custom plugin to gate tool calls. They have permission config, which I'm using to block file reads, but the orchestrator is allowed to use specif
9. 549.▲
   by verdverm 2mo ago

   1. I don't have the agents write the code handsfree, I'm involved in the process so I know what's actually there 2. review/validate, yes, this still requires you getting into the guts, ai does not write good code, HITL i
10. 550.▲
    by verdverm 2mo ago

    I could argue that one implementation adding these things does not make a std library. You lose portability and fracture the ecosystem, which is probably one of their intentions
11. 551.▲
    by verdverm 2mo ago

    I called them "artificially incessant" after I watched our PR orchestrator agent use subagents to work around permissions to read files, despite instructions that explained the intentional restrictions. I've since added more
12. 552.▲
    by verdverm 2mo ago

    Ant is also buying politicians and attempting regulatory capture, not things I'm keen to have corporations doing
13. 553.▲
    by verdverm 2mo ago

    this right here, call out your project best practices and peculiarities in AGENTS.md I suspect eventually this will move to fine-tuning, there are multi-tenant LoRAs on the horizon that will let us keep paying per-token on FTs
14. 554.▲
    by verdverm 2mo ago

    The point is that the ease of workaround is besides the point, this is about Anthropic's approach. Lest we forget the silent nerfing, the silent prompt modifications, refusals of valid user prompts, watermarking, and lies about distill
15. 555.▲
    by verdverm 2mo ago

    maintenance, being a good player in the ai ecosystem
16. 556.▲
    by verdverm 2mo ago

    people use it because it does in fact make a sizable difference in experience and outcomes
17. 557.▲
    by verdverm 2mo ago

    Fireworks and OpenCode both have excellent open model offerings prefer open companies to closed ones
18. 558.▲
    by verdverm 2mo ago

    it's always been this way with Claude/Ant though, not something that changed This is more SV modus operandi
19. 559.▲
    by verdverm 2mo ago

    sanitation services are very different from convenient consumerism, so I won't take it as a reason to accept drones Do you really think Amazon, of all companies, will spend more money to make quieter drones? (that will likely still be
20. 560.▲
    by verdverm 2mo ago

    we could argue about license semantics all day, either way Google is not being a good player
21. 561.▲
    by verdverm 2mo ago

    I suspect people will complain about the noise if an incident doesn't ground it first. I certainly do not want the propeller whining around
22. 562.▲
    by verdverm 2mo ago

    it drones on and on, it lacks a coherent train of thought I assume it is agent written, slop regardless I'd be very hesitant to call a piece of writing about Ai "watertight" when the author starts with > I want to start o
23. 563.▲
    by verdverm 2mo ago

    I heard it described as an inversion of the OPEC cartel. China can now influence prices as much as they can/could, up to a point. They prepared in Feb by buying 200M barrels, as if they had uncannily high confidence Trump was going to
24. 564.▲
    by verdverm 2mo ago

    some real slop, don't waste your time
25. 565.▲
    by verdverm 2mo ago

    I personally like the way Scion breaks it down into - model - harness (tools/config) - agent (live/running) https://googlecloudplatform.github.io/scion/concepts/ Scion allows for multiple configurations
26. 566.▲
    by verdverm 2mo ago

    Almost every harness has an SDK now, it seems part of the "mvp" at this point, both open and closed source https://learn.chatgpt.com/docs/codex-sdk https://code.claude.com/docs/en/ag
27. 567.▲
    by verdverm 2mo ago

    Does it have anything co-designed around Vercel infrastructure? This is what happened to NextJS and why I will likely never touch Vercel open source again
28. 568.▲
    by verdverm 2mo ago

    yes, in the tools I use, each (sub)agent will also load the same files (automatically, root is always loaded, a nested AGENTS.md is read if the dir or a peer file is touched) yes, they are useful, mainly in that they shorten the context gat
29. 569.▲
    by verdverm 2mo ago

    well I generally agree, pretty much every other harness supports looking for the various alternatives and uses them I see this obstinance as an intent signal and one of the smaller bullet points I have for avoiding Ant
30. 570.▲
    by verdverm 2mo ago

    choose not to support them in return if it is a real problem otherwise a simple symlink from AGENTS.md -> CLAUDE.md works well enough disclaimer, I only use open weight models and open source harnesses so have no stake in this either way