File pilot seems to use an immediate mode UI to be responsive and uses 10% cpu even when idle or minimized. I'm not sure that's worth it.

The training data is not so specifically filtered at least in pre training. The point is to give them as much world knowledge as possible

The OP is saying maybe that was a bad idea. I tend to agree given how badly these companies manage to sanitize outputs.

Yeah that was my first thought. Use a TUI that abuses the terminal to make a experience poorly recreating an actual GUI, don't be shocked if it doesn't turn out very efficient

I actually like the XS one. It's broken, but it's very aesthetic

I just admit, I didn't expect this to make it to HN

AI written/edited front page.

A passion project and you can't even trust yourself to write the front page.

Made me lose interest.

Ah, so certain, and so wrong.

(Author here) All 19 edits to the file floppy.md over the past few months are in my local git repo, and all the people who proof-read different versions of it would be happy to attest. The site is statically generated from templates and content created by me.

The manifesto is written in my attempt at the same style as others I've read, like the one for Small Games (as in scope, not file size): https://twitter.com/gingerbeardman/status/196592155732860525...

The small games manifesto has no LLM tells.

If yours really is fully handwritten, then you've been reading so much AI writing you're mimicking a lot of its patterns, which is sad to see.

So you're now not so certain. That's progress, I guess. You seem to want to tell me what I've been doing and how I've been doing it, which is a bit odd.

Maybe you could look at using some AI assistance to provide some real 'useful' feedback to the OP about what made you lose interest, as your false confidence in some opaque 'AI-detection' heuristic is lazy and only harms individuals who are trying to find effective ways to share information.

The AI tells are quite literally what made me lose interest. They help me filter out dozens of AI slop posts daily so I'll be keeping them thanks. They're lazy by design. You can generate a good looking post in 30s, I can't afford to spend 5 mins reading each one on its merits.

> summarizing web pages

For summarizing creative writing, I've found Opus and Gemini 3 pro are still only okay and actively bad once it gets over 15K tokens or so.

A lot of long context and attention improvements have been focused on Needle in a Haystack type scenarios, which is the opposite of what summarization needs.

I think it's less about the extremity of evil and more about lacking the means to get rid of it in a more civil manner.

It is tho’ Naive not to

Interesting. Would be able to release it on api or weights so we can use it outside the context of your application?

@bolwin we will definately do that in future. Still in alpha and making updates and scaling it up but hopefully in next 3-6months we will have an API available

> The numbers came from the same project and the same prompt across versions.

I'm pretty sure the tester checked. If the request format is the same (which it is, given it uses the same as Anthropic's stable public API) and the same prompt/messages then bytes will correlate pretty well.

The prompt may be the same, but the project context would have have surely changed. User prompt itself is unlikely to be ~200KB.