Hacker News Reader

devanshpon A global workspace in language models

https://claude.ai/share/2b0f85a2-e7b8-4f62-91a0-eca61bdeabec

Fable 5 on low gets the answer with web search turned off, one-shot!

devanshpon Regex Chess: A 2-ply minimax chess engine in 84,688 regular expressions

This is absurd. I did not realize you could do nearly this much computation in regex.

devanshpon We collected 10k hours of neuro-language data in our basement

Cool post! I'm somewhat curious whether the data quality scoring has actually translated into better data; do you have numbers on how much more of your data is useful for training vs in May?

devanshpon An Interactive Guide to the Fourier Transform

It's quite interesting that our ears implement a better-than-Fourier-like algorithm internally: https://arxiv.org/pdf/1208.4611

devanshpon Building the heap: racking 30 petabytes of hard drives for pretraining

we have 6 months of experience operating thousands of physical disks in datacenters now! it's about a couple hours a month of employee time in steady-state.

devanshpon Building the heap: racking 30 petabytes of hard drives for pretraining

no, we wanted to be conservative by depreciating somewhat more aggressively than that. we have much closer to 5% yearly disk failure rates.