ramoz

15 ms·

1. 331.▲
   Show HN: OpenCode plugin for interactive plan annotation(github.com)

   3 points by ramoz 10mo ago|0 comments
2. 332.▲
   by ramoz 10mo ago

   > Takeaway 3c: Experienced developers disagree about using agents for software planning and design. Some avoided agents out of concern over the importance of design, while others embraced back-and-forth design with an AI. Im in the back-
3. 333.▲
   by ramoz 10mo ago

   Not really. But, You can go viral again with a "Coding Agents with memory build better software using less tokens" showcasing how you benchmarked a "twitter rebuild" - 1. Setup Claude Code to build some layers of the sta
4. 334.▲
   by ramoz 10mo ago

   Inter has also become my default.
5. 335.▲
   by ramoz 10mo ago

   If markdown in a git repository isn’t good enough for collaboration, then why would any plugged in abstraction be better? You imply you have a solution for current wholistic state. For this you would need a solution for context decay and re
6. 336.▲
   by ramoz 10mo ago

   I found this happen less often if the task is a part of the plan. It typically gets in a cycle habit of editing code and updating the doc
7. 337.▲
   by ramoz 10mo ago

   I wasn’t able to get my point across. But I completely agree
8. 338.▲
   by ramoz 10mo ago

   The more concerning algorithms at play are how they are post-trained. And the then concern of reward hacking. Which is what he was getting at. https://en.wikipedia.org/wiki/Reward_hacking 100% - we really shouldn&#x27
9. 339.▲
   by ramoz 10mo ago

   Part of my initial pitch was to automate linting. Interesting insight on the stop loop. Ive been wanting to explore that more. I think there is a lot to be gained also with llm-as-a-judge hooks (they do enable this today via `prompt` hooks)
10. 340.▲
    by ramoz 10mo ago

    Great advise. For large plans I tell the agent to write to an “implementation_log.md” and make note of it during compaction. Additionally the agent can also just reference the original session logs.
11. 341.▲
    by ramoz 10mo ago

    I agree with this take. Esp with the simplicity of /sandbox I created the feature request for hooks so I could build an integrated governance capability. I don’t quite yet think the real use cases for hooks has materialized. Through a
12. 342.▲
    by ramoz 10mo ago

    I’m interested to see your setup. I think planning is a critical part of the process. I just built https://github.com/backnotprop/plannotator for a simple UX enhancement Before planning mode I used to write plans to a
13. 343.▲
    by ramoz 10mo ago

    My somewhat naive heuristic would be that memory abstractions are a complete mistep in terms of optimization. There is no "super claude mem" or "continual claude" until there actually is. https://backnotprop.c
14. 344.▲
    by ramoz 10mo ago

    Here is a simple skill (markdown instruction only) that instructs a nice ripgrep approach - with utility of discovering current session. https://github.com/backnotprop/rg_history
15. 345.▲
    by ramoz 10mo ago

    I struggle with these abstractions over context windows, esp when anthropic is actively focused on improving things like compaction, and knowing the eventual* goal is for the models to yave real memory layers baked in. Until then we have to
16. 346.▲
    Show HN: Interactive plan annotation and sharing for Claude Code(github.com)

    4 points by ramoz 10mo ago|0 comments
17. 347.▲
    by ramoz 10mo ago

    They acquired in order to have an ASICs competitor to Google TPU.
18. 348.▲
    by ramoz 10mo ago

    If using Claude Code, a simple hook can govern `browser_navigate` (mcp) A custom sh script or something for whitelists would take ~5min to setup. For more robust governance (many policies), you can write Rego using https://github
19. 349.▲
    by ramoz 10mo ago

    This is also what I’ve been doing. I can’t find the exact video but this is another good example: https://www.tiktok.com/@thinkwithv/video/7580186972208024863 The comments are usually insightful. Even in that vide
20. 350.▲
    Color Schemes(observablehq.com)

    4 points by ramoz 10mo ago|0 comments
21. 351.▲
    by ramoz 10mo ago

    big reach
22. 352.▲
    by ramoz 10mo ago

    > Histotripsy generally seems to stimulate an immune response, helping the body attack cancer cells that weren’t targeted directly by ultrasound. The mechanical destruction of tumors likely leaves behind recognizable traces of cancer pro
23. 353.▲
    by ramoz 10mo ago

    I just saw a video of non-technical person describing how they use claude code to automate various workflows. They actually tried vscode and then the desktop gui. Yet they preferred the CLI because it felt "more natural" With agen
24. 354.▲
    by ramoz 10mo ago

    I haven't come across a case where it has used the LSP yet. Opus 4.5 is fairly consistent in running QA at proper times. Lint checks and all are already incorporated into a standard & native processes outside of IDE. I think lookup
25. 355.▲
    by ramoz 10mo ago

    > Our expectation is that combining multiple approaches—a defense-in-depth strategy—can help cover gaps that any single method leaves exposed. Implement hooks in codex then.
26. 356.▲
    Show HN: Agent Skill turns existing filesystem into Claude's memory(github.com)

    4 points by ramoz 10mo ago|0 comments
27. 357.▲
    "it runs on your computer"(backnotprop.com)

    5 points by ramoz 10mo ago|0 comments
28. 358.▲
    by ramoz 10mo ago

    A sufficiently sophisticated agent, operating with defined goals and strategic planning, possesses the capacity to discover and circumvent established perimeters.
29. 359.▲
    File System as Claude Code's Memory(backnotprop.com)

    2 points by ramoz 10mo ago|0 comments
30. 360.▲
    by ramoz 10mo ago

    https://x.com/giansegato/status/2002203155262812529/photo/1 https://x.com/METR_Evals/status/2002203627377574113 > Even Google's cheaper Gemini 3 Flash model seems to be