x.com/hilbertspaess/status/2108996523040366647

This sentence makes a universal factual claim ("everywhere"), but the available published evidence includes clear counterexamples.

1) Anthropic explicitly says it has not observed sustained AI-attributable doubling in its own progress.
On October 2, 2026, Anthropic's Transparency Hub stated: "We have not observed a sustained doubling in the pace of our AI progress attributable to AI." That directly contradicts the idea that uplift is doubling on a 3–6 month cadence everywhere, because Anthropic is one of the most relevant places such a trend would need to hold.

2) METR's public productivity-uplift survey does not show 3–6 month doubling either.
In METR's May 11, 2026 survey of 349 technical workers, respondents retrospectively estimated about 1.3x value uplift in March 2025, 2x in March 2026, and 2.5x for March 2027. That is growth over roughly year-long intervals, not a doubling every 3–6 months.

3) METR also reported that Anthropic itself did not think its large internal uplift estimates were translating into anything like repeated doublings of overall progress.
In METR's May 19, 2026 Frontier Risk Report, METR wrote that Anthropic had shared an internal survey finding a geometric-mean 4x productivity uplift, but Anthropic also said it didn't trust that estimate much and believed they were seeing less than 2x overall increase in the pace of progress. METR summarized Anthropic's view as suggesting that even a 4x increase in labor input might map to only about 1.7x output.

Because the claim says uplift is doubling every 3–6 months everywhere, a single well-documented exception is enough to falsify it. Anthropic's own published statements are such an exception, and METR's broader survey data also does not fit that doubling claim.