i doubt there has ever been an organization as intellectually principled—and as willing to follow an idea all the way down—as @OpenAI.
in my two years here, i’ve watched us repeatedly follow the implications of the bitter lesson—even, at times, to a fault (e.g., our fixation on RL and competitive coding arguably came at the expense of interactive/agentic coding for a while). the underlying instinct has been consistent: bet on methods that scale, resist hand-engineered dead ends, and let evidence win.
we compete the same way: build the best thing, offer it to everyone, and keep moving the frontier. we haven’t selectively cut off access to our models, even when a game-theoretic reading of the playing field might have counseled otherwise.
my favorite part is our obsession with “pushing things left” into the model. i get to work with frontier companies to close the capability overhang in their products and chart the jagged frontier. every workaround and every product failure points to an unclaimed capability.
it is an extraordinarily fun intellectual loop to be part of.
Good Sunday. Reset has been propagated to accounts and we landed some fixes to usage for things mentioned yesterday as issues we found. You should feel a positive difference. More to come tomorrow and will keep communicating.
@rezoundous I can't afford Terra bro u r talking about ultra? As soon as submit prompt my eyes always turn to speed meter with each percentage my heart break