TL;DR
The short version
Breakthroughs that look sudden are the payoff of decades of slow groundwork — AI took roughly 50 years to become an "overnight" success. The visible breakthrough is the tail end of a long curve, not a standing start. A leader's job is to separate what's genuinely arriving from what only feels imminent, and to expect the lag.
The throttle on a technology is often not compute but the slowest thing it's fused to. The VR Kevin Kelly first saw in 1987 isn't much better today — just vastly cheaper — because the bottleneck is eyes, weight, and focus, not processing. Anything bound to the human body or to institutions moves at the speed of biology and habit, not silicon.
Drawn from Every's AI & I — Kevin Kelly interviewed by Dan Shipper (published 29 Jul 2026, recorded ~April 2025). The dated history verifies; the forward-looking claims are Kelly's predictions.
The 50-year overnight success
AI's sudden arrival rode five decades of quiet work — the field dates to the 1956 Dartmouth workshop, the deep-learning breakthrough to around 2012. What looked like a standing start was the steep end of a curve that had been climbing, mostly invisibly, the whole time.
You could say VR is still waiting for its LLM moment.
Kevin Kelly
VR is the counter-case that proves the rule: demoable since 1987, still stalled almost 40 years later. The same logic predicts embodied robots arrive far later than the hype implies — anything fused to the human body moves at the speed of biology, not compute.
Predictions are easy; true predictions are hard
The operator skill isn't forecasting — it's forecasting accurately, which means separating genuine arrival from felt imminence. Kelly's own misses (VR too fast, eBay, Bitcoin-as-micropayments) are the evidence, and the honesty is the method: a forecaster who won't list their misses isn't giving you a usable signal.
Amara's Law is the compressed form — we overestimate a technology's impact in the short run and underestimate it in the long run. The rigorous version is Carlota Perez's model: a technology's installation (arrival, hype, financial capital) precedes its deployment (productive payoff diffused through the economy) by years. That gap is where the "50-year" lag lives.
Expect the lag — but don't mistake it for a reason to wait
For a futurist the lag is philosophical; for a product leader shipping an AI feature this quarter it's a budget problem — real ROI deadlines, not 50-year arcs. Both are true at once, and holding both is the discipline.
