Malleable software in the age of LLMs (2023)(geoffreylitt.com)

73 points by tosh 2 years ago | 26 comments

Enough AI copilots, we need AI HUDs(geoffreylitt.com)

902 points by walterbell 1 year ago | 249 comments

Geoffrey Litt's blog post arguing that AI should be designed as heads-up displays augmenting perception rather than copilot agents, drawing on Mark Weiser's 1992 critique.

Understanding is the new bottleneck(geoffreylitt.com)

401 points by sebg 2 months ago | 221 comments

Geoffrey Litt's written version of an AI Engineer conference talk arguing developers still need to understand agent-written code, with ideas for understanding it efficiently.

Bicycle(ciechanow.ski)

2962 points by todsacerdoti 3 years ago | 336 comments

Bartosz Ciechanowski's interactive explainer article on how bicycles work, covering the physics of riding, steering, and balance with 3D demonstrations.

A hackable AI assistant using a single SQLite table and a handful of cron jobs(geoffreylitt.com)

722 points by stevekrouse 1 year ago | 159 comments

Geoffrey Litt's blog post describing Stevens, a family AI assistant built with one SQLite memories table and cron jobs hosted on Val.town.

AI doesn't generate working products, that's still your job(weeraman.com)

210 points by smckk 2 months ago | 195 comments