bill
bill
@brob.bsky.social
dilettante with a vague mental map of vast territories brob.town
168 followers369 following225 posts
started dabbling in are.na the Pinterest for Hipsters and the lack of feed ranking is already killing me
wake up babe, new andy matuschak talk just dropped
⭐ New talk! Watch / read / visually skim here: andymatuschak.org/tat Coding agents might help us finally break out of two cages: the app model, which traps computing in one-size-fits-all silos; and programming as a specialization, which has crowded out cultures of imagination and domain insight.
bill reposted
even worse, engineers may soon be expected to have “taste” some notes on how we might do this here: brob.town/notes/2024-1...
we settled the “should designers write code” debate years ago. now it’s time for retribution. engineers: you’re learning color theory, type hierarchy, grid systems, copywriting, and navigation design. welcome to hell
to the fine folks @anthropic.com idk if you gather preference data for post training from consumer chat logs but if so may I suggest grepping for “You’re right, I was over-engineering it. Here’s the simplified version”
LLMs are currently bad at very basic questions involving standard musical notation and theory. Trying to convince @komaniecki.bsky.social to help build some evals for this
this thread articulates one of the most cogent explanations of the different senses of the term “agent” i’ve seen yet
"No one knows what the hell an AI agent is" is true, but this article stops at the easy part: techcrunch.com/2025/03/14/n... People are currently using "agent" in at least three very different ways:
some friends and i are imagining what a pedagogically motivated IDE for nontechnical folks might look like and i will be updating my design doc to quote this skeet, ty @tedunderwood.me
This new world of ours rewards being a dilettante with a vague mental map of vast territories.
the client is open source. so one might dig through that to understand what signals the Discover algo might be using beyond the core interaction data (likes, follows, graph structure) to make recs…
* while much of bluesky's code is open-source, Discover (the default algorithm) is not. but you can create and subscribe to tens of thousands of alternate feeds and even remove the default Discover from your app — algorithmic choice!
also i don’t know whether the official bsky client logs this stuff or not - just observing that one could investigate and find out, or rub the magic lamp and bug @pfrazee.com to see if he appears
the client is open source. so one might dig through that to understand what signals the Discover algo might be using beyond the core interaction data (likes, follows, graph structure) to make recs…
* while much of bluesky's code is open-source, Discover (the default algorithm) is not. but you can create and subscribe to tens of thousands of alternate feeds and even remove the default Discover from your app — algorithmic choice!
the client is open source. so one might dig through that to understand what signals the Discover algo might be using beyond the core interaction data (likes, follows, graph structure) to make recs…
* while much of bluesky's code is open-source, Discover (the default algorithm) is not. but you can create and subscribe to tens of thousands of alternate feeds and even remove the default Discover from your app — algorithmic choice!