Hacker Newsnew | past | comments | ask | show | jobs | submit | lenadroid's commentslogin

Turns out you can build a neural network with your hands. No code or GPUs, just switches, knobs, and a needle that tells you what it sees.

This is a perceptron. The knobs in the middle are the weights. To train it, you turn them by hand, one small nudge at a time, until the needle points the right way.

A similar process is happening inside LLMs. Just with a trillion more knobs and a GPU doing the turning.

Akamai’s CTO Dr. Robert Blumofe built this on his workbench and open-sourced everything — schematics, PCBs, Gerber files, 3D-print frames, bill of materials, the works.

Full 20-minute video walkthrough (super clear and actually fun to watch): https://www.youtube.com/watch?v=PSqP73T0g_M

Complete build guide + all files: https://github.com/rdb64-hobbies/Perceptron/blob/main/BUILDI...

Only basic soldering required. Bonus 3D printer friendly.

This is such a beautiful way to actually get a feel for how neural nets work. Has anyone here built one yet? Would you try it?


I spent 3 days at AI Engineer World’s Fair in San Francisco, with three thousand of the world's best AI engineers, Fortune 500 CTOs, and founders.

I chatted with engineers, architects, and founders in companies like Open AI, Pydantic, Microsoft, etc. to get their thoughts on some of the relevant questions on the future of work with AI.


"the theory behind failure detection, a couple of real-world examples of how the mechanism works in a real cloud - on Azure. Even though this article includes real-world applications of failure detection within Azure, the same notions could also apply to GCP, AWS, or any other distributed system."


many significant updates are coming for EH, definitely a thing to watch


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: