orsorna

It is partially the result from chilling effects of hosting AI generated code on your public portfolio. Not worth the ire.

No because you likely do not drench yourself in hubris daily.

I am wondering what use people get out of this. What I like about pi is that, using proprietary harnesses to begin with, I already somewhat know what kind of features I use and features I wish I could use. Asking pi to self modify itself (with a smart enough model) is a one time cost for a fully vetted feature that I personally approve. Do users of OMP find all of its additions useful?

>He's never going to be hired as CEO by anybody for the rest of his career

I'd eat my shorts if this was true. But even then he is likely set for life with his wealth, pre Meta offer.

Fallacy that only serves corporate interests, as he points out.

Oh, I thought it was some weird branding thing. So then I looked it up.

"System One" and "System Two" were coined in some pop science book...so back to its usage being a marketing ploy.

For anyone who was piqued by the title but couldn't stomach the unedited LLM blogpost, you can read the similarity AI-genned docs for Laya directly: https://github.com/NandhaKishorM/laya

It was a breath of fresh air for me because it was nice going to a model that is "smart enough".

>The trillion dollar question is how you do this, if your employees do not care (they are optimising for salary & time spent not code quality)

Sounds like there is a compensation problem then.

You definitely need more RAM if you are not satisfied with small context windows, especially if the weights take a large % of the total memory to boot.

Is it that silly? You could run multiple 27B models in parallel.

Context https://www.businessinsider.com/mark-zuckerberg-used-a-boost...

I don't do anything fancier than inference, and I only use llama.cpp, which supports rOCM. I've had few issues; most GGUFs I download work right out of the box. Nearly any popular model has a quant that just works. But as you can see I don't use my GPU for anything weird or nonstandard.

7900 XT is a sleeper card. When I initially bought it, it was priced at the lowest wattage per $ per GB VRAM (not normalized for token speeds...) Although I ended up swapping for the XTX because that 4GB means everything in just increasing the context window. At 8bit KV my window is over 200k, and although qwen3.8 loves vomiting out tokens as part of its reasoning chain I trust it enough to get assigned tasks done eventually, which I could not say of any model before its release.

This is just blogspam, right? The author even brags about hitting the HN front page in such a short amount of time, in a previous post. All he talks about is his agent (that we are supposed to "know"? arrogant). Not a lot of discussion from him about his results.

All signs point to these donations being strictly ideologically motivated, and not because of the alleged merits Omarchy totes. I could further argue how sad it is that efforts like these receive tons of money versus struggling FOSS projects, all for a strictly non-technical reason.

There is also an open source tvOS client called Swiftfin. The experience isn't as good though.

You can also transcode your files to formats that the free version of Infuse supports if you are a miser.

What investor concerns prompted writing this one?

A Faustian library, containing every book known but you aren't allowed to read them all without being evicted from the library.

I guess I couldn't believe it. The LLM copy on that homepage is unbearable and amateurish.

Whatever twitter.now is which is neither.

Who in their right mind would take that site seriously?

Does applying to a job immediately really increase your chances that much?

Documentation is only useful for agents if the source is unavailable, and if you are trying to save on tokens and trust that the docs are accurate.

If none of that is relevant to you, or rather accuracy is your concern, it's much better just to search the source code for your answers, from how to set up the application to configuration questions to troubleshooting. Bonus points if you strip out all the comments programmatically via linting before you even let the agent view the code, to avoid possible context poisoning.

Not really any indication that you are collecting data for your own usage. No source code at least?

On that note, would it really be hard to use my own deployment of your app to check against the datasources your deployment uses?

The reason he said that is because he's a rube.

Is the demand low because the capital requirements to perform the research aren't there? Excluding wages...granted, for an average research project I don't know what % goes to wages and admin overhead. Probably the bulk?

[flagged]

Really hate this characterization, although to be fair I love the ops/infra side of things more. I've used to set up so many parts of my homelab manually, but now I can just get it all done. I've spent countless hours maintaining a Rpi server at the infancy of my interest in computing. Truly a lot of it is maintenance and busywork tasks. I am having more fun than ever. Perhaps some people like those busywork sysadmin tasks and gain enjoyment of the homelab hobby just from that...honestly I can't say I really understand it.

Same reason why grifters adopted web3 when Web 3.0 was already an established term.