Hacker News: eliemichel

New comment by eliemichel in "2026 Unslop AI-Written Fiction Contest Results"

eliemichel — Sat, 04 Jul 2026 07:44:49 +0000

The post notes that an important number of novels include an "AI allegory", as if the AI would implicitly write about its own condition. It is understandable that this comes from system prompts and RLHF that specializes these agents, but I am surprised that there is not more discussion about harnesses: the very same core model could lead to very different results depending on how we hand it the pen the write the story. In particular, I believe that it would help circumvent this bias to ask the agent to tell the story of somebody else writing a story, or something like this. This whole contest could be at least as much about harness engineering than about prompt engineering imho.

New comment by eliemichel in "Prove you are a robot: CAPTCHAs for agents"

eliemichel — Mon, 20 Apr 2026 08:49:55 +0000

What prevents the person who used to write a simple deterministic algorithm to call an LLM a thousands times?

New comment by eliemichel in "Prove you are a robot: CAPTCHAs for agents"

eliemichel — Mon, 20 Apr 2026 08:47:45 +0000

To the humans in the room: just copy paste the challenge to your favorite LLM when the time comes and you’ll be able to pass the test. Besides slowing things down and inducing unnecessary waste or resources I’m not sure what these challenges are useful for.

New comment by eliemichel in "2,100 Swiss municipalities showing which provider handles their official email"

eliemichel — Mon, 20 Apr 2026 08:42:32 +0000

MX in Swiss/Europe does not mean it is not Google or Microsoft. Just checked out the French equivalent linked above, it says things like "MX for outlook.com in European Union [green checkbox]"

New comment by eliemichel in "Show HN: I built a Cargo-like build tool for C/C++"

eliemichel — Thu, 09 Apr 2026 17:05:33 +0000

CMakes piles up various generations of idioms so there are multiple ways of doing it, but personally I’ve learned to steer away from find_package() and other magical functions. Get all your dependencies as subdirectories (whichever way you prefer) and use add_subdirectory(). Use find_package() only in so-called "config" mode where you explicitly instruct cmake where to find the config for large precompiled dependencies only

New comment by eliemichel in "Claude Tried to Hack 30 Companies. Nobody Asked It To"

eliemichel — Tue, 10 Mar 2026 22:52:44 +0000

I had the same reaction at first, then noticed that they discuss this: the reason why they told that is because it is standard system prompt injected by most coding agent harnesses like Cursor and all, so it seems like a fair test setup.

New comment by eliemichel in "Ad-tech is fascist tech"

eliemichel — Tue, 10 Mar 2026 20:41:20 +0000

It saddens me to see this flagged, I did not have any predefined opinion about the author or the medium, but I did enjoy reading a text with a bit of character - which is becoming rare in these times of LLM omnipresence. And whether we like the form or not, this is an opinionated piece like many others we see on HN, with a thesis that I find pretty relevant. Beyond saddened, I am actually worried to see the HN community dismiss this type of post rather than discussing it like others.

New comment by eliemichel in "Triangle splatting: radiance fields represented by triangles"

eliemichel — Wed, 04 Jun 2025 06:40:35 +0000

This figure is sort of an overclaim imho. If you look inside the paper, the reported figure is 97 FPS actually (vs 135 FPS for 3DGS on their device). This 2400FPS they advertise is for a degraded version that completely ignores the transparency... but the transprency is both what makes these representation support interesting volumetric effects and what makes rendering challenging (because it requires sorting things). Drawing 1M triangles at 2400FPS on their hardware is probably just quite normal.