OpenAI выбирает деньги: почему из компании уходят топовые ученые

OpenAI is no longer the cozy laboratory where scientists could spend years contemplating the nature of mind. Recent news of a mass exodus of researchers confirms it: the company has definitively transformed into a product-manufacturing machine. This time, the suitcases were packed by vice president of research Jerry Tworek, head of model policy Andrea Vallone, and economist Tom Cunningham.

This is not merely a rotation of personnel, but a genuine tectonic shift in Sam Altman's strategy. The root of the problem lies in resource reallocation. While fans await GPT-5, OpenAI's internal processes increasingly resemble a conveyor belt.

Leadership has decided that now is not the time for long-term and risky experiments. Instead, all forces are directed at the "front" — at ChatGPT. One must acknowledge that OpenAI is in the position of a defending leader.

Google and Anthropic are snapping at its heels, and every week of delay costs the company millions of dollars in potential revenue and market share. In such conditions, fundamental science becomes an unaffordable luxury that managers ruthlessly cut. Jerry Tworek's departure is particularly painful.

He was one of those who stood at the origins of the company's key achievements. When such people leave, they take with them not only knowledge, but that very culture of "crazy ideas" that made OpenAI great. Now the company focuses on iterative improvement of existing models.

This is the classic trap of a successful startup: when you become too big, you stop taking risks and start optimizing. Researchers accustomed to changing the world do not want to spend time polishing the user interface or fixing minor bugs in chatbot logic. Andrea Vallone, responsible for model policy, was also a key link in the dialogue between engineers and regulators.

Her departure against the backdrop of global disputes about AI safety looks at least troubling. It may mean that commercial pressure has become so strong that questions of ethics and long-term risks have taken a backseat. If previously OpenAI tried to balance between "safe" and "fast," now the scale has clearly tipped toward speed.

The situation reminds one of the history of many great IT corporations of the past. First, they attract the world's best dreamers, then transform them into well-paid cogs of the system. For the industry, this is an important signal: the golden age of open research in private companies is coming to an end.

Now breakthroughs should be watched for either in university laboratories or in new startups created by the very refugees from OpenAI. Altman is building a business empire, and in that empire, scientists seem to be becoming increasingly cramped. We are watching as OpenAI definitively loses its identity as a scientific organization and becomes "Microsoft 2.

0" — efficient, profitable, but stripped of its former magic. The bottom line: OpenAI has definitively chosen the path of commercialization. Will the company be able to maintain its leadership without its chief visionaries, or are we heading for an era of stagnation where instead of new architectures we get only cosmetic updates to ChatGPT?