Hacker News: elcomet

New comment by elcomet in "Hammerspoon"

elcomet — Sat, 14 Mar 2026 10:55:47 +0000

Could you share your config?

What does a lean proof prove?

elcomet — Tue, 16 Dec 2025 09:44:11 +0000

Article URL: https://supaiku.com/what-does-a-lean-proof-prove

Comments URL: https://news.ycombinator.com/item?id=46286605

Points: 5

# Comments: 1

New comment by elcomet in "Reservoir: Smart electric water heater"

elcomet — Sat, 15 Nov 2025 07:56:03 +0000

The heat pump will pull heat from inside the house? This sounds terrible for efficiency in winter, as you will need to reheat the room

New comment by elcomet in "Reservoir: Smart electric water heater"

elcomet — Sat, 15 Nov 2025 07:54:54 +0000

Yeah this sounds disingenuous. I can also see my heater to 70C if I want, that does not increase its size...

New comment by elcomet in "Gemini 3.0 spotted in the wild through A/B testing"

elcomet — Fri, 17 Oct 2025 07:02:13 +0000

I don't understand all the hype for generating SVG with LLM. The task is not really useful, doesn't seem that interesting in single shot as it's really hard, and no human could do it (it would be more useful if the model has visual feedback and could correct the result).

And also, since it becomes a popular task, companies will add the examples in their training set, so you're just benchmarking who has the better text to SVG training set, not the overall quality of the model.

New comment by elcomet in "ProofOfThought: LLM-based reasoning using Z3 theorem proving"

elcomet — Sun, 05 Oct 2025 08:18:33 +0000

I'd say that thoughts and reasoning are two different things, you're moving the goalpost.

But what makes the computer hardware fundamentally incompatible with thinking? Compared to a brain

New comment by elcomet in "ProofOfThought: LLM-based reasoning using Z3 theorem proving"

elcomet — Sat, 04 Oct 2025 19:51:42 +0000

How can you know?

New comment by elcomet in "Blender 4.5 LTS"

elcomet — Fri, 03 Oct 2025 05:24:22 +0000

Can you explain more? Which things are impossible in blender

New comment by elcomet in "Progressive JSON"

elcomet — Sun, 01 Jun 2025 08:43:59 +0000

Too many acronyms, what's FE, BFF?

New comment by elcomet in "Firefox tab groups are here"

elcomet — Tue, 29 Apr 2025 16:46:41 +0000

Arc browser unifies the tabs and bookmarks in a very clever way.

New comment by elcomet in "I've been using Claude Code for a couple of days"

elcomet — Sun, 09 Mar 2025 16:25:45 +0000

I'm wondering if you can prompt it to work like this - make minimal changes, and run the tests at each step to make sure the code is still working

New comment by elcomet in "Microsoft begins turning off uBlock Origin and other extensions in Edge"

elcomet — Fri, 28 Feb 2025 08:10:28 +0000

If it's chromium based, they will need to remove manifest v2 at some point to stay close to the upstream version.

New comment by elcomet in "Mistral Small 3"

elcomet — Thu, 30 Jan 2025 21:23:12 +0000

You can offload tensors to the cpu memory. It will make your model run much slower but it will work

New comment by elcomet in "Prime Numbers So Memorable That People Hunt for Them"

elcomet — Tue, 21 Jan 2025 18:05:13 +0000

I can't tell if this is a joke if if you're serious

New comment by elcomet in "Why making friends as an adult is harder"

elcomet — Tue, 24 Dec 2024 17:09:36 +0000

Book clubs, art clubs, movie club... Lot of options

New comment by elcomet in "Ask HN: Who's building on Python NoGIL?"

elcomet — Mon, 23 Dec 2024 13:19:14 +0000

What do you mean? Async/await uses threads

New comment by elcomet in "Offline Reinforcement Learning for LLM Multi-Step Reasoning"

elcomet — Mon, 23 Dec 2024 11:11:50 +0000

Multi-step reasoning means that the LLM is giving a question (maths here), and generating an answer that consists of many intermediate words, before returning the solution. Here, we don't want to tell the LLM how to solve the problem word-by-word. We want to tell it at the end, "correct" or "incorrect", and have the model learn on its own to generate intermediate steps, to reach the solution.

That's typically a setup where RL is desirable (even necessary): we have sparse rewards (only at the end) and give no details to the model on how to reach the solution. It's similar to training models to play chess against a specific opponent.

New comment by elcomet in "OpenAI O3 breakthrough high score on ARC-AGI-PUB"

elcomet — Fri, 20 Dec 2024 23:26:54 +0000

He means 10 humans voting for the answer

New comment by elcomet in "Gukesh becomes the youngest chess world champion in history"

elcomet — Thu, 12 Dec 2024 18:33:12 +0000

Those are tree search techniques, they are not metrics to assess the "human" complexity of a line. They could be used for this purpose but out of the box they just give you winning probability

New comment by elcomet in "Raspberry Pi boosts Pi 5 performance with SDRAM tuning"

elcomet — Tue, 03 Dec 2024 13:34:33 +0000

I'm wondering why there isn't a cheapest alternative to the pi without those features that most find useless. Or a more powerful for the same price