1+ day, 7+ hour ago  (1113+ words) I've been quietly obsessed with Universal Transformers for almost six years. The question UT was asking " should the network choose how much computation to spend per input? " was always the right one. It is being asked again, under different names....

1+ week, 2+ day ago  (1817+ words) I expected the most-downloaded skills on Claw Hub to be the best-written. They're not. When I had Gemini grade the top 100, it marked down 84% of them. One popular skill lost 30 points. The tool is kura (open source). The archive is…...

1+ mon, 3+ day ago  (1536+ words) It started with a simple idea: what if you played as a storm cloud terrorizing a city? That's roughly what my son and I told Claude. No design document, no technical spec " just the concept. Claude took that idea and…...

1+ mon, 2+ week ago  (757+ words) I've been thinking about a gap between what code agents do and what they understand. They write functions, fix localized bugs, pass benchmarks " but ask them to modify something that touches six modules in a real codebase and things fall…...

2+ mon, 2+ week ago  (793+ words) There's a legend that once, John Carmack (creator of id Software, to whom we owe thanks for Wolfenstein, Doom, and Quake) asked Ilya Sutskever what he should read to master deep learning. And Sutskever gave him a list of forty…...

8+ mon, 2+ week ago  (499+ words) Tversky proposed an alternative: a feature matching model where similarity is a function of common and distinctive features. Despite its psychological plausibility, this model relied on discrete set operations, making it incompatible with the differentiable, gradient-based optimization that underlies modern…...