[echelon]

Then start using DeepSeek.

Don't use distilled little RTX models on your frankensteined home PC like a 0.00001%er who misses the ergonomics Claude Code solves. That's a "Year of Linux on Desktop 2010" grade failure waiting to happen.

Rent cloud instances and spin up thick model weights and contribute to the open source infrastructure for making this easy for everyone to use.

The hyperscalers should be eaten by cheap, competent, cloud-based open source.

Be the change you want to see.

[esafak]

Which GPU cloud instance are you using?

[echelon]

Lambda, RunPod, and I've used GCP and AWS instances and clusters too.

I'd prefer if we had a more open version of all of this. That's where open source will matter.

[moosedev]

Please don’t post AI slop to HN. It’s against the guidelines.

[echelon]

> Please don’t post AI slop to HN.

??? - My comment doesn't even look like an LLM wrote it.

> It’s against the guidelines.

To be clear:

I wRoTe

- E V E R Y - w o r d -

by [picture ASCII art of "human hand" here]⠀

Do not accuse me of using an LLM as a substitute for thinking.

I do not use LLMs to write any of my Hacker News comments, and I never have.

I have used LLMs to help me search for information, but I have never once in my life posted the outputs of an LLM into a post verbatim as my own. That's laziness.

I'm not dismissing LLMs here. They produce better writing than a lot of humans. They are fantastic tools for constructing media, for getting work done, and for furthering your own capabilities.

But I curate my thoughts and craft my arguments. That sort of wanton dismissal of my comments is worse than using an LLM as a substitute for thinking. That's real "human slop".