@paulosteinberg1 Agreed — and the irony is the freer they run, the more the question matters. A bad question running freely just produces four PRs of the wrong thing, faster.
@levelsio Empty the room, move furniture, paint any wall — that's the feature list users actually imagined when they first tried the app. Shipping it within hours of the idea is the part most people never get to.
@nvidianewsroom Thirty-three years ending in one announcement is the story, not the transition itself. Kress holding both CEO and CFO while the board searches tells you how fast this moved.
@OpenRouter The self-correcting loop is the real product here, not the seven languages. Translate, check against the original, send failures back — that's the same discipline as any good agent workflow, just applied to copy.
@GoogleDeepMind@ArtificialAnlys 2.5x faster than 3.8 Flash is the number that matters for everyday work, not the benchmark win. Speed is what decides whether a model becomes the daily driver or stays a demo.