Discussion 2 years ago: https://news.ycombinator.com/item?id=39720388

The similiar course in

- Stanford: https://web.stanford.edu/class/cs208e/cgi-bin/main.cgi/

- MIT https://ocw.mit.edu/courses/6-080-great-ideas-in-theoretical...

- Harvard https://beta.my.harvard.edu/course/COMPSCI1/2026-Spring/001

So, three-quarters of the people remained. That's a good result.

I'm a serial entrepreneur from China, and I've been here for 14 years now. HN is still the website I check every single day.

Truly grateful to everyone in this community.

Do you have any plan to turn it into a full book—maybe called Grokking Linear Algebra ?

Lol. Maybe! I did enjoy writing Grokking Algorithms, but writing a full book is a real commitment. That one took me 3 years.

Actually Z.ai is a spinoff of Tsinghua University and one of the first China labs open sourcing its own large models (GLM released in 2021) . https://github.com/THUDM/GLM

With a little search you can find it's a laboratory within the CS department of THU. It's a fairly large lab though, not those led by just one or two professors.

You can run GLM-4.5-Air 4bit on Macbook Pro M4 Max 128GB for local vibe coding: https://x.com/ivanfioravanti/status/1949854575902523399

Cool to know about, but I'm starting to get the impression that watching an LLM spit out hundreds of lines of rudimentary css is like the new buying a car to drive to the grocery store. When we look back on how we've trained ourselves to forget how to tolerate walking or design neighborhoods for livability, will we be happy we sacrificed the insects and air quality to do so?

Not intended to be dismissive though, I do it, it's just getting kind of icky to me.

Also no fun, I didn't learn how to code to be the project manager for an artifically intelligent machine.

You can choose to write the juiciest parts yourself and let the AI deal with grunt work.

AI writes code, I drink coffee and watch it write code.

And what do you see yourself doing in 5 years?

Seems like you're being intentionally facetious, but the best output that I've seen from any of the llms is stuff that we did a ton of 5 or more years ago, which is great because we're still doing a lot of that, it just sometimes helps get it done a bit faster after going through the steps to learn how to do it manually and correctly.

Coding till my eyes hurt everyday. Cuz I love it.

Still saying “we must treat the LLM as a junior”

Fixing Vibe Code!

Can it run in thinking mode locally, where it runs commands and edits files?

Thank you! just found the 4bit (60gb) on LM Studio

CogAgent: Open source alternative from China.

Paper： https://arxiv.org/abs/2312.08914

Code： https://github.com/THUDM/CogAgent

Model: https://huggingface.co/THUDM/cogagent-9b-20241220

Click3: open source alternative for Mobile operator

Code: https://GitHub.com/BandarLabs/clickclickclick

See also the book or the course by Harry R. Lewis at Harvard:

- Ideas That Created the Future https://www.amazon.com/Ideas-That-Created-Future-Computer/dp...

- "Classics of CS" https://canvas.harvard.edu/courses/34992/assignments/syllabu...

I am curious whether this theory can explain some phenomena of large language models.