Hacker News: dymk

New comment by dymk in "Run Kimi K3 using 29 GB of RAM at 0.50 tok/s"

dymk — Sat, 01 Aug 2026 04:05:18 +0000

That doesn’t mean “use fewer thinking tokens”. It might use more as it mulls over how to make its response concise.

New comment by dymk in "Agent Skill to Force Docs in ASD-STE100 Simplified Technical English"

dymk — Fri, 31 Jul 2026 22:20:00 +0000

Projects like SkillOpt[1] are pretty good evidence that the ceiling for skills isn't "describe it in a sentence and let the LLM generate it". There's value in iterating on a skill and figuring out what the weak parts are. At that point, sharing the skill can indeed be useful. Just look at superpowers, GSD, etc. They're 'just' skills (and a few bash scripts), but the text in them is useful enough to share.

[1] https://microsoft.github.io/SkillOpt/docs/guideline.html

New comment by dymk in "Agent Skill to Force Docs in ASD-STE100 Simplified Technical English"

dymk — Fri, 31 Jul 2026 06:39:22 +0000

No, it was a viral tweet from last week. They’ve been verbose since GPT-4o.

New comment by dymk in "Agent Skill to Force Docs in ASD-STE100 Simplified Technical English"

dymk — Fri, 31 Jul 2026 06:15:58 +0000

The point of STE100 is to avoid ambiguity like this, though.

New comment by dymk in "Agent Skill to Force Docs in ASD-STE100 Simplified Technical English"

dymk — Fri, 31 Jul 2026 06:14:36 +0000

Precommit hook

New comment by dymk in "Agent Skill to Force Docs in ASD-STE100 Simplified Technical English"

dymk — Fri, 31 Jul 2026 06:12:34 +0000

You can memorize a book too but that doesn’t mean they’re not useful

New comment by dymk in "The AI Aesthetic"

dymk — Fri, 31 Jul 2026 04:02:41 +0000

I’ve noticed much the same thing. Adding rules for ruthless minimalism and no anticipatory features helps, if context hasn’t degraded too far. Still, I tend to ask it to remove cruft much more often than I ask it to add new UI.

New comment by dymk in "A Trampoline"

dymk — Wed, 29 Jul 2026 22:00:18 +0000

Backyard trampolines without nets are particularly dangerous. A gymnasium or trampoline park with in-ground trampolines are my go-to if someone wants to learn how to do tricks.

New comment by dymk in "Discovering Cryptographic Weaknesses with Claude"

dymk — Tue, 28 Jul 2026 23:28:43 +0000

This feels like a subtle attempt to start an argument about Star Wars being sci-fi or not.

New comment by dymk in "A Discord Alternative – Echoed"

dymk — Mon, 27 Jul 2026 23:37:40 +0000

Front page says 150MB RAM, About page says 90MB.

Whole site is full of LLM-isms (em-dashes, stilted 'X. Y. Z.

.' constructions).

It's so boring. It says "We put one prompt worth of effort into our landing page". Just like all the others.

"No ads · No tracking · No ID verification · Free forever" - why would anyone believe this for a closed source application that, presumably, you want to monetize? Servers and developers aren't free, unless it's open source and/or self hosted.

New comment by dymk in "How is the Bun rewrite in Rust going?"

dymk — Mon, 27 Jul 2026 15:19:22 +0000

He had schizophrenia.

New comment by dymk in "London Gatwick has launched a robotic airport parking service"

dymk — Sun, 26 Jul 2026 19:20:20 +0000

A train system is a much safer bet than praying traffic won't be bad

New comment by dymk in "Kill The Cookie Banner"

dymk — Sun, 26 Jul 2026 15:15:30 +0000

Then sign in, thats enough signal for them to track you

New comment by dymk in "The growing vigilante movement to knock out Flock surveillance cameras"

dymk — Sat, 25 Jul 2026 22:10:15 +0000

They certainly have knives, go look at what a tiger is sporting on his paws

New comment by dymk in "Stolen Buttons"

dymk — Sat, 25 Jul 2026 21:13:56 +0000

What a button look like to you? Blue roundrect sure says button to me.

New comment by dymk in "Claude Cookbook"

dymk — Fri, 24 Jul 2026 16:45:21 +0000

Then let's not pretend what software engineers have been doing for the last 50 years is engineering, either. But what else would you call it?

New comment by dymk in "Claude Cookbook"

dymk — Fri, 24 Jul 2026 16:44:32 +0000

Understanding how to write good prompts (prompt engineering) is still very much a relevant skill if you want to effectively use LLMs. Harnesses aren't magic.

New comment by dymk in "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"

dymk — Tue, 21 Jul 2026 23:32:57 +0000

The weights are the output artifact, the training corpus and system are the source.

New comment by dymk in "Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling"

dymk — Tue, 21 Jul 2026 06:09:11 +0000

If that’s the case I still don’t get it. 640k was enough in 1981, same as how 150 (or 9k) tok/sec would suffice for 2026

The comment seems like the nerd equivalent of 6 7

New comment by dymk in "Agent swarms and the new model economics"

dymk — Tue, 21 Jul 2026 00:13:59 +0000

This post describes ideas found in existing agent orchestrator systems since they've existed. You'll find the same "top level agent that breaks a problem down into spec / plan / implement / verify" pattern in get-shit-done or superpowers.

Is the original part the VCS that works at 1000 commits/second? Or spending a bunch of money comparing models?