Was chatting to friends tonight, a couple of lawyers and an engineer, and they brought up using Claude. I asked them what they thought of Fable 5. Blank stares, they hadn't heard of it. X is a bubble
@jarradgrigg ๐I feel like the pitch deck has always been overhyped, but I still like the idea of a short supporting memo to tell the story, but lead with product
@jarradgrigg Nah I was in South West. A lot more property development than I remember. Very keen on land + build down there but prices looked pretty crazy
Wonder if the reason Three.js is such a popular use case for Fable is because the algo keeps feeding me similar examples or because Fable 5 seems particularly good at spatial reasoning, visual feedback loops, and long horizon agentic iteration
Used this as an excuse to rebuild the beach scene for The Good Stuff podcast
@jarradgrigg Ideally. First attempt was rigged box-part bodies with procedurally-authored voxel heads but looks incredibly lame. Need to work out the best way to do this. Any ideas?
Firmly in the camp that FIFA is rigging this World Cup for Messi. Watched a compilation of the fouls not given to Egypt. Just wild.
That said, not parking the bus with a two goal lead and ten mins to play is incredibly naive
Lot of pundits eating their words about Bellingham. I watch this guy every week as a Madrid fan, he'll retire the greatest ever English player and it won't be close
Hands down the best podcast studio. Really enjoyed hanging out with @Pete_Winn and @andymdavid to discuss AI privacy and what to do about AI agents that are such big gossips!
We also explored what the tech landscape looks like in a few years: more software or less software?
Great chat with @Justinmoon for the Big Episode 50 on the Good Stuff Podcast!
Nice deep dive into Justin's AI psychosis and lessons learned from building about 500 projects in the last year
Crazy that we're already up to 50 episodes
https://t.co/XKBNHDB4SA
@VullinghsJ Unsure if youโre already doing this, but a knowledge graph as a memory layer could be useful here for context selection, and maybe eval, since you can inspect whether the agent surfaced the right entities, relationships, and evidence