OpenAI Making Choices
OpenAI chose to cut Sora. An excellent example of strategy in practice.
In a recent interview, OpenAI president Greg Brockman explained why the company chose to cut its video generation programme (Sora) and double down on the GPT reasoning model tree. His explanation had nothing to do with market sizing or SWOT-models. It was a single sentence, in classic engineer lingo:
"The sum of random vectors is zero. But if you align your vectors, you can go in a direction."
Think of it this way: if ten people in a room each push a table in a different direction, the table doesn't move. But if they all push the same way, it shifts in a meaningful way. That is what Brockman is describing. It is a great illustration of what we mean when we say strategy is choice.
The Decision
OpenAI had two genuinely promising technology branches: the GPT reasoning series and the Sora world model series. Both were working. Both had real potential. However, pursuing both made it impossible to accumulate the compute, focus, and institutional momentum required to move decisively in either direction. So OpenAI made the harder choice: cut Sora, double down on GPT.
The decision was not a judgement that video generation is unimportant. It was a recognition that effort scattered across competing priorities does not add up. It cancels out. This is what Roger Martin means when he says strategy requires integrated choices. Choices should amplify the others, not merely coexist with them.
The Cost of Keeping Options Open
Most organisations resist this kind of choice because keeping options open feels safe. If we maintain both bets, the thinking goes, we can be right about either one.
But there is a cost that is easy to undercount: diffusion. Every resource committed to one path is a resource not compounding on another. Over time, spreading effort across multiple directions does not reduce risk. It guarantees that nothing goes far enough to win. Sooner or later, you'll be outcompeted.
The Harder Question
The question Brockman's decision forces on any leadership team is not "what should we prioritise?" That framing implies a ranked list of activities you intend to pursue in sequence. The harder question is: which of our current directions are we willing to stop, so that the ones we keep can actually go somewhere?
OpenAI's choice is a useful mirror for any leadership team. The instinct to keep options open is natural, but it carries a hidden cost: diffusion. When effort is spread, nothing compounds fast enough to win. The organisations that move are the ones that pick a direction, accept what that means giving up, and push together.
We help leadership teams work through their strategic choices and test the critical assumptions that underpin them. If this resonates, get in touch.