Curated tech Twitter/X posts on AI, startups, and product development—signal without the doomscroll, updated on weekdays.

Focus

Pulse Checkwhat people say about tools, with sources Productsnew launches makers are posting Profilesthe accounts we watch Topicsclusters the feed keeps returning to

//

Articleslong-form writeups pulled from the feed Time Travelreplay any day in the archive Threadsthe Dan's Dimes thread archive

//Zeitgraphs//Newsletter//About//

Post

Press Space to continue

Keep reading

Related tweets

›

›

›

More from this author

›

›

›

Zeitgraphs·Dan's Dimes·LGTM

More From Us·Zeitgraphs·Dan's Dimes·LGTM

Built byIndie Thinkers

GLM-5 Architecture Uses Multi-Head Latent Attention | Tech Twitter

GLM-5 Architecture Uses Multi-Head Latent Attention

Press Space to continue

The weights are out! Here's the GLM-5 architecture comparison. GLM-5 is: - bigger than its predecessor (mainly more experts) but has rel. similar active parameter counts - uses multi-head latent attention - uses DeepSeek Sparse Attention

Content

Keep reading

Related tweets

›

›

›

More from this author

›

›

›

Previous tweet Next tweet More from @rasbt