Genetic algorithms were supposed to deliver human-level AI by 2025. Transformers did — on time, and nothing like the path Kurzweil described.
In 2005 Kurzweil named the decade, the substrate, and the architecture. In 2024 he quietly moved every date ten years. But the deeper question is whether ‘nanobot’ was the wrong word all along — and CRISPR, base editors, and LNPs are fulfilling the original vision in a body Kurzweil didn’t name.
Character.AI hit 20M users three years early; brain nanobots didn’t show up. Kurzweil’s VR, BCI, and mind-upload predictions scored against 2026 reality.
Kurzweil’s 2005 warfare chapter got the shape right and the institutions wrong: FCS died, smart dust didn’t arrive, and Ukraine’s $500 quadcopters are the force he described.
Kurzweil said global brain observation would arrive in the 2020s. In September 2025, 621,733 neurons across 279 brain areas, in Nature. Twelve neuroscience predictions, scored.
Kurzweil said we wouldn’t need to map every connection in the brain. The connectomics field disagreed and just spent a decade doing exactly that.
The cubic-millimeter human connectome arrived. The nanobots did not. A scorecard on twelve Kurzweil predictions about brain reverse-engineering.
Ten ALS patients walk around with a bloodstream-delivered brain reader. It just isn’t a nanobot.
Kurzweil’s 2005 nanotech predictions: where laboratory demos were real, what got built anyway, and which substrates actually won the decade.
Kurzweil promised molecular assemblers by the 2020s. In 2024 he quietly moved the date to the 2030s. Twelve nanotech predictions, scored against 2026 reality.