This is cool. Meta describe how they operate infra and deploy services.

cacm.acm.org/research/met...

we know it's AGI when AI models naming becomes intelligible

Finally found a concise and comprehensible overview of DeepSeek v3

epoch.ai/gradient-upd...

A recipe for improving SQLite concurrent write performance:

1. PRAGMA journal_mode=WAL;
2. PRAGMA busy_timeout=5000; or higher.
3. PRAGMA synchronous=NORMAL;
4. Keep write transactions small.
5. Avoid write-after-read transactions.
6. Use app-level locking if needed.

Is it just me or bluesky's Suggested Accounts shows only politicians?

A great lecture on container networking: how to implement bridge networking on a host, across hosts, etc.

www.youtube.com/watch?v=6v_B...

Yes! a 15-min-read post to catch up on all experiments, reproductions, explorations around DeepSeek R1

We've summarized everything we did & saw in this first week since DS came to light

Results/code/dataset/experiment => if we missed one, share & we'll update

Link: huggingface.co/blog/open-r1...

Companies out here be like…

DeepSeek Debates: Chinese Leadership On Cost, True Training Cost, Closed Model Margin Impacts

semianalysis.com/2025/01/31/d...

There is nothing easier than writing something you want to write, and nothing harder than writing something you don't want to write.

My new post explores how SQLite handles concurrent writes

tenthousandmeters.com/blog/sqlite-...

And if you wanted an overview of how SQLite works, here it is, a lecture by the SQLite creator www.youtube.com/watch?v=ZSKL...

SQLite can give you an excellent write performance, e.g. ~70k inserts/s. Concurrent writes are ok, but if you have too much concurrency, performance drops, e.g. with 128 threads inserting, it's ~3k inserts/s . There's a trick: use a mutex to sync threads, same 128 threads give ~50k inserts/s.

not a huge fan of air cooling

Let's get it started!