Will the US government build AGI? (“The Project”, graded)

What he predicted

One of Situational Awareness's boldest claims is institutional, not technical: as AGI approaches, Aschenbrenner argued, the US national-security state would inevitably step in and run a government-led AGI effort — explicitly analogous to the Manhattan Project — by roughly 2027 or 2028. He treats this as close to inevitable once the stakes become legible.

Where it stands in mid-2026

The evidence is mixed, which is why this is graded Open rather than wrong. On the "moving toward it" side: national-security attention to AI has grown markedly — export controls, security requirements at the frontier labs, and growing defense interest. On the "not happening" side: there is no formal, centralized government AGI project of the kind the essay describes. Frontier development remains led by private labs.

What would settle it

The prediction's deadline is 2027/28, so it has not yet resolved. It would be graded fulfilled if the US government launches a centralized, Manhattan-Project-style AGI effort by then — and Wrong if that deadline passes with frontier development still firmly in private hands. Until then the trend is partial: more state involvement, but no Project.

Why this prediction matters

The Project is downstream of the open-source verdict. The essay's case for government takeover rests on AI being concentrated and controllable. If capability instead diffuses cheaply — as the open-source verdict suggests it has — the strategic logic that motivates a centralized Project weakens.

Frequently asked questions

Unresolved as of mid-2026. Aschenbrenner predicted a Manhattan-Project-style US government AGI effort by 2027/28. National-security involvement is growing, but no formal project exists yet, so the prediction is graded Open.

Aschenbrenner's term for a predicted centralized US government AGI effort, analogous to the Manhattan Project, that he expected the national-security state to launch by roughly 2027/28.

Not in a centralized way. As of mid-2026, frontier AI development is still led by private labs, though government attention — export controls, security requirements, defense interest — has grown significantly.