Post

Luis Villa
@lu.is

Oh mannnnnnn

conputer dipshit@davidcrespo.bsky.social

the talk doesn't quite put it this way but I had a revelation reading it: it's completely backwards to be generating plans and code alone and then collaborating on the review. we should be doing nearly all our collaboration on the planning. we do this in various ways already but it should be More

Luis Villa
@lu.is

Some takeaways from reading this amazing talk; I should: - demand many more diagrams of state machines - I should use Lavish a lot more - I should reevaluate my current inclination to talk less, code more; but how to talk-collaborate effectively is going to be hard

Eleanor Saitta
@dymaxion.org

Don't just demand digrams for state machines, demand declarative specifications for them that are converted to procedural code *deterministically*. Nothing else works reliably, including human post-hoc review of code.

1 repost2 likes

Eleanor Saitta
@dymaxion.org

There should be state machine generator libraries for most languages (parser generators, too). The ideal is formal correctness guarantees, but those are sadly still rare. Haven't been tracking what's current, though.