Hacker News: anuramat

New comment by anuramat in "The Text Box Is the Command Line of AI"

anuramat — Fri, 29 May 2026 09:55:16 +0000

so, AR? how is AI relevant here?

also, since when is CLI outdated?

New comment by anuramat in "Claude Code – Everything You Can Configure That the Docs Don't Tell You"

anuramat — Fri, 29 May 2026 08:15:20 +0000

still, far more effective than "NEVER FUCKING GUESS"

New comment by anuramat in "Claude Code – Everything You Can Configure That the Docs Don't Tell You"

anuramat — Fri, 29 May 2026 08:05:40 +0000

fyi you can re-enable it with `{ "showClearContextOnPlanAccept": true }` in ${CLAUDE_CONFIG_DIR:-$HOME/.claude}/settings.json

also find `"disableAutoMode": "disable"` useful, since I'm typically switching between yolo and plan

New comment by anuramat in "Claude Code – Everything You Can Configure That the Docs Don't Tell You"

anuramat — Fri, 29 May 2026 07:53:23 +0000

in my experience, "undocumented tricks" break as often as documented features

like when they removed "clear context and execute plan" option after releasing 1M opus because "context window is not a problem anymore"

New comment by anuramat in "Sam Altman Won in Court Against Elon Musk. But, We All Lost"

anuramat — Wed, 27 May 2026 11:03:18 +0000

I'd say "plagiarism machine" implies that

New comment by anuramat in "Sam Altman Won in Court Against Elon Musk. But, We All Lost"

anuramat — Fri, 22 May 2026 15:42:01 +0000

you know you can set the temperature to zero and get exactly that, right?

New comment by anuramat in "Sam Altman Won in Court Against Elon Musk. But, We All Lost"

anuramat — Fri, 22 May 2026 15:38:17 +0000

why would anthropic do that?

New comment by anuramat in "Sam Altman Won in Court Against Elon Musk. But, We All Lost"

anuramat — Fri, 22 May 2026 15:37:22 +0000

> wrong 10% of the time

that doesn't sound nearly as bad as you think it does; I don't see how ethics are relevant here either, unless oil is also somehow a scam

regardless, one must be delusional to deny the fact that it's useful tech

"but they're evil" is not an argument

New comment by anuramat in "Sam Altman Won in Court Against Elon Musk. But, We All Lost"

anuramat — Fri, 22 May 2026 15:28:00 +0000

you think it should be deterministic?

New comment by anuramat in "No Slop Grenade"

anuramat — Thu, 21 May 2026 12:31:00 +0000

now I know what to call it, thanks

New comment by anuramat in "Alternatives for the EDIT tool of LLM agents"

anuramat — Wed, 20 May 2026 04:24:03 +0000

so, hashline?

New comment by anuramat in "Show HN: Id-agent – Token efficient UUID alternative for AI agents"

anuramat — Tue, 19 May 2026 19:32:18 +0000

was wondering myself, just tried comparing to petnames crate -- gets you about 2 tokens per word on average

not that anyone should ever care; typos in random-looking ids are very real but already covered by human readable ids

besides, this is for a specific tokenizer

New comment by anuramat in "Rewrite Bun in Rust has been merged"

anuramat — Thu, 14 May 2026 18:58:11 +0000

I think their point was that the project is complex, with the implicit assumption that the complexity is to a large degree inherent.

Even if it's mostly accidental, and the code is overengineered slop (which it is), the system being able to decompose a problem and deliver something is impressive in terms of stability: it wasn't sucked into rewriting everything from scratch every time it would run into issues, it didn't have infinite subagent recursion with a one-agent-per-line type workflow, etc.

New comment by anuramat in "Fake building: Claude wrote 3k lines instead of import pywikibot"

anuramat — Tue, 12 May 2026 04:33:35 +0000

There are actually multiple ways to completely eliminate the cognitive load of any sort of activity, so please, don't settle on half measures

New comment by anuramat in "Ask HN: What are you working on? (May 2026)"

anuramat — Mon, 11 May 2026 03:55:49 +0000

vicode -- TUI coding agent written in Rust, with tabs/subagents running in worktrees on top of fuse-overlayfs: create/fork tabs to work on multiple features/implementations, while subagents work in parallel without conflicts; additional lowerdir with bindfs mounts lets agents share the compilation cache, so that `cargo check` doesn't take minutes

since it's all just mounts, vicode works as a worktree manager as well: select a vicode tab (which sets cwd to the corresponding worktree with OSC7), open a new terminal tab/window, and run claude/codex inside

disclaimer: unstable, linux-only (mac build WIP, no overlayfs), some modules were vibecoded (grep for `SLOP`)

https://github.com/anuramat/vicode

New comment by anuramat in "Statecharts: hierarchical state machines"

anuramat — Mon, 04 May 2026 16:33:01 +0000

you forgot to replace the em dashes :^)

New comment by anuramat in "Ask HN: Will local models on normal hardware ever compete?"

anuramat — Mon, 27 Apr 2026 05:06:40 +0000

I'm pretty sure an average machine will always be less capable than a datacenter, the rest depends on your definition of super capable

New comment by anuramat in "A 3D Body from Eight Questions – No Photo, No GPU"

anuramat — Sat, 25 Apr 2026 23:51:34 +0000

makes sense in a philosophical debate or when you're talking to your confused grandparents, but does anyone on hn not know how LLMs work, at least on the level of "tokens, matrices, data, sgd"?

otherwise, that reminder must imply that people do know how it works, and yet they still ascribe to these models some property like qualia, i.e. something other than "being able to turn english into code and compute into shareholder value";

but then if you disagree, why even mention it in the first place? do atheists randomly proclaim "btw god isn't real!" in unrelated conversations with strangers of unknown religious beliefs?

New comment by anuramat in "Lambda Calculus Benchmark for AI"

anuramat — Sat, 25 Apr 2026 16:56:58 +0000

"almost as good as opus at writing python/js/... when given a spec" might be enough for a lot of people, especially if its 10x cheaper

New comment by anuramat in "Show HN: A Karpathy-style LLM wiki your agents maintain (Markdown and Git)"

anuramat — Sat, 25 Apr 2026 16:49:04 +0000

been thinking the same, but I imagine you could explicitly separate notes and slop, eg something as simple as cron job that goes through all your notes and creates a PR if there's some easy win: typos, inconsistencies, tags, etc

I've been coding like this lately: if I'm too lazy to review a new non-critical section/unit tests, I'll mark it as `// SLOP`; later, if I have to, I'll go through the entire thing, and unmark

shitty tests are better than no tests, as long as you your expectations are low enough