mrkn1

very indeed. not sure whether i'm buying intelligence or approval, but definitely paying with data and money.

Vibe coder sounds more legitimate than bug breeder.

Imagine being so good at your job that someone else greatest intellectual achievement is looking at your presentation.

A very academic way of saying We have all the data in the world, except the data we actually need. You ask a company how something works and they say:

"Oh, theres a process."

Great. Where is it documented?

"in a spreadsheet."

Which spreadsheet?

"Well, theres the old spreadsheet, new spreadsheet, and the spreadsheet that Karen maintains."

Where's Karen?

"She left in 2021."

good point, what do you suggest then?

Clickable link: https://kouh.me/ytfeed

Clickable link: https://kouh.me/ytfeed

Yeah quality has gotten worse. Also, some mods love power a bit too much.

There is a Reddit Search, but in my experience is very underwhelming, so not sure that counts.

Thanks

"Validation of this claim is impossible". How so? There's no network traffic.

If you press F12 on your browser, open the Network tab, you can watch every single request that page makes. Hope it helps!

Please send your example to hi@kouh.me I'll investigate

Agreed. This video goes over how it works https://m.youtube.com/shorts/m_GcDH1vOl0

White noise does nothing to privacy. Stop the misinformation.

It's free and unlimited, Pangram is paid. Try a few examples to see performance and speed. It's faster than Pangram.

Thanks for checking it out. Can you send me the samples to hi@kouh.me to reproduce?

This video explains how it works: https://m.youtube.com/shorts/m_GcDH1vOl0

Theres a CLI tool for this gap. textsnap on github runs OCR locally/offline, with no arg it reads an image from clipboard and copies the extracted text back to it. Might not solve the App Store install restriction, but could cut the friction if you can get it running via something like iSH or a Linux vm

you can transcript fast for free on cpu with yapsnap (github repo)

7/8 result is already Lean-verified, yet experts are spending their time on the unverified 11/12 paper because it’s shorter and the only one a human edited

Worth noting there’s also textsnap if you want something that’s actually local end-to-end via ONNX on CPU, no GPU, no API key, nothing leaves your machine. https://github.com/kouhxp/textsnap

Yup, it's q8. I would try q4 and q2 and if you are still unhappy with MB size, limited vocab.

this is such a poor model, I've tried it before. Try yapsnap it has better word error rate and runs super fast on CPU. https://github.com/kouhxp/yapsnap/

you can use textsnap on CPU (i made it). https://github.com/kouhxp/textsnap

yapsnap is faster on CPU and uses a zipformer https://github.com/kouhxp/yapsnap/

> once celld has confirmed they're no longer needed

Isn't that the same check then act race in a new spot, and is it covered by an invariant?

Great guide! It works, made a 500MB model for CPU based off qwen 3.5 0.8B. It plays maze games like Pac man https://github.com/kouhxp/gutsy

subword tokenizers were never causal in the first place, so BPE was peeking at future bytes all along! TIL

Thank you, that makes perfect sense. I would likely go the clf route once I have enough data for a task.

Cool project too. If you are looking for a 500MB instead of gigabytes, with evals on Jevbench that you can run fast on CPU check out gutsy.

https://github.com/kouhxp/gutsy

thanks!