Sign in

shiminsky

@shiminsky.bsky.social
166 followers 251 following 53 posts

Software Developer, Dog Dad, Beginner Gardener, Host of the Artificial Developer Intelligence (ADI) Podcast www.adipod.ai -- em-dashes my own

PostsRepliesMedia
shiminsky @shiminsky.bsky.social · 05/08/2026
Tomato garden finally coming online.
A bowl of tomatoes of varying size
000
shiminsky @shiminsky.bsky.social · 10/06/2026
me this morning:
A confused Snowy
010
shiminsky @shiminsky.bsky.social · 10/06/2026
Was working with Fable to last night to design a chakram / frisbee inspired flying machine. The physics seems to checkout -- but Im no aerospace engineer.
A 3d rendering of a chakram-based flying machine
100
shiminsky @shiminsky.bsky.social · 24/04/2026
Think of it as a "delusion index', the delta between a model's self judgement and it's peers score. Weaker models tend to have more delusion since they are unable to differentiate complex work. Gemini is the real outlier here -- Opus and GLM 5.1 also had surprisingly 'objective' self scores.
A bar chart showing the difference between a model's score of its own output versus the average peer models' scores
000