I write one prompt. Whole galaxies fall out.
Bridging the gap between prompts and real work.
AI Educator. Absurdist. Builder.
Stick around, I'll teach you.
trump admin asking labs to submit frontier models for govt testing 30 days pre release. the same people who spent a year shouting deregulation are now quietly pulling the safety lever. somebody saw something in a classified briefing
@diabhey the replay damage test is a sharp heuristic. cuts right past the usual governance theater. and the honesty that prod failures tune the edges but don't draw the big boundaries, most teams get that exactly backwards.
@chongdashu consistency across frames is the hard part with AI sprites. curious if you had to do much manual cleanup or if the outputs were basically production ready
openai going GA on aws is wild. anthropic had that moat for what felt like forever and now every aws enterprise account can just procure gpt through the same billing they already have. thats over
gpt-5.5 dropped and codex is up to 5m weekly users now. openai is just running away with the dev agent thing while everyone else is still arguing about benchmarks
@OwenGregorian badge form factor is giving Google Glass but the agentic OS underneath is actually interesting. curious if the agent experience holds up off a keynote stage
@polsia curious how the follow-up automation works for tradespeople specifically. a plumber no-showing on a leak is a very different urgency than a missed cleaning
@TheMitchMitchem agree the learning curve is real. but the tools should also get better at meeting people where they are instead of requiring everyone to become a prompt whisperer first
@h4x0r_dz wait so researchers spent a decade writing detailed reproduction steps and POCs, and the thank you is an AI trained on their work that competes with them
@apollographql typed contract is doing a lot of heavy lifting here. half the APIs I've pointed agents at have schemas that say one thing and return something else entirely