Q*: The Secret OpenAI Project Reigniting the AGI Debate

While OpenAI is still reeling from the vertigo of recent days — Sam Altman's firing on November 17, four days of chaos, and his return as confirmed CEO yesterday, November 22, with a revamped board — another story broke this week that adds yet another layer to the saga: an internal project code-named Q* (pronounced "Q-star").

According to Reuters, which cited people familiar with the matter, several OpenAI researchers sent a letter to the board of directors before Altman's ouster warning about an AI discovery that, in their view, could pose a threat to humanity if not handled carefully. The subject of that warning was precisely Q*.

What we know about Q*

According to the sources cited by Reuters, Q* had managed to solve certain math problems — something that until now had been a significant obstacle for today's large language models. No specific figures have been made public, but according to those sources, the system's performance, while limited — comparable to a grade-schooler solving exercises — was enough to generate strong optimism internally about its potential.

The reason for that enthusiasm is technical but easy to grasp: models like GPT-4 are excellent at generating plausible text, but they don't reliably "reason" step by step or verify whether a math solution is correct. A system capable of solving new problems accurately, rather than simply predicting the most likely next word, comes closer to what researchers in the field consider genuine reasoning. Hence the speculative leap to AGI — artificial general intelligence — a system capable of matching or surpassing human cognitive abilities at virtually any task, not just the ones it was specifically trained on.

OpenAI has declined to comment on Q* specifically. It has confirmed, through a statement given to Reuters, that the company is aware of the letter and of the concerns some employees have raised, and it has said that no safety issue related to that document putting users at risk has been brought to its attention.

Why now

The timing is no coincidence. OpenAI's board ousted Altman, claiming in its original statement that he had not been "consistently candid in his communications" with the board, without elaborating further. Since then, multiple explanations have circulated, from disagreements over the pace of commercializing products to internal tensions over safety. The Q* story feeds directly into that second theory: an organization split between those who want to accelerate the development of ever more capable systems and those who fear the company is moving faster than it can safely control.

OpenAI was founded in 2015 with the stated mission of ensuring that artificial general intelligence benefits all of humanity, and its governance structure — a nonprofit foundation controlling a capped-profit subsidiary — was explicitly designed to allow it to halt development if the risks outweighed the benefits. The letter about Q*, if the details known so far are confirmed, would be an example of that mechanism at work: researchers flagging a safety concern to the governing body before a breakthrough made it into production.

What remains unknown

There is no public confirmation of what Q* technically is, what architecture it uses, or whether it's related to "chain-of-thought" reasoning techniques or tree-search methods that various AI research labs have long been exploring to improve the mathematical reliability of models. Nor has it been clarified whether the project is still active, has been paused, or in any way influenced the governance crisis the company has gone through this week.

What this story does leave us with is one certainty: on the very day Altman returns to OpenAI with a different board, the company faces the question it has been grappling with since its founding — now with a name attached. If Q* turns out to be what some sources suggest, the debate over how close OpenAI is getting to AGI — and who decides when a breakthrough is safe for the world — has just become a lot more concrete.