UpTrajectory Review

Kevin Kelly, the founding executive editor of Wired, spent much of 2016 telling anyone who would listen that AI was about to swallow the economy, and that the rational response was to become the person who directs the machines rather than the person the machines replace. A decade on, a retrospective posted at taylor.town grades those predictions against what actually happened, and the exercise is more useful than most anniversary journalism because Kelly was right early, loudly, and in public — which makes his misses as instructive as his hits. The piece is circulating on Hacker News, which is exactly the audience that spent those ten years living through the transition Kelly forecast.

For a small-business operator, the retrospective lands differently than it does for a software engineer. Most of you did not retrain as prompt engineers in 2017; you were busy making payroll. Kelly's core claim — that AI would not eliminate work but restructure it around people who could orchestrate automated systems — has quietly become true in the least glamorous corners of the economy. The contractor who now runs takeoff estimates through a vision model, the clinic that automated intake paperwork, the retailer using generative tools for product copy: these are the 'centaurs' Kelly predicted, half human judgment and half machine throughput. The retrospective's value is confirming that this was not a temporary hack but the stable shape of white-collar work.

What is genuinely contested in the piece is not whether AI mattered — that argument is over — but whether Kelly's optimism about augmentation versus displacement holds at the frontier of current models. The original 2016 framing assumed human oversight would remain the bottleneck; the 2026 reality is that agentic systems increasingly plan, execute, and verify multi-step work with minimal supervision. We are skeptical of any retrospective that grades decade-old predictions without acknowledging that the last eighteen months broke the curve the previous eight years traced. Kelly predicted the centaur; he did not predict the autopilot.

The second-order effect worth watching is credential deflation. If a decade of AI adoption validated Kelly's advice to 'learn the machines,' the next decade may punish the operators who treated that as a one-time retraining rather than a permanent operating posture. The businesses that thrived were not those with the most sophisticated AI but those with the shortest feedback loop between trying a tool and changing a workflow. That advantage compounds: early adopters accumulated process knowledge that late adopters cannot buy off the shelf, which is why two businesses in the same trade can now have wildly different cost structures.

Our advice: read the retrospective not as prophecy scored but as a checklist. Identify one workflow in your operation that still assumes a human must manually transfer information between systems — that is the task the next wave of agentic tools will eliminate first. Pilot it this quarter, measure the hours, and decide whether to redeploy that labor toward revenue or let it attrit. Kelly's 2016 bet was that adaptability, not scale, would be the moat. Ten years of evidence says he was right, and the operators who act on that now will not need another retrospective to tell them so.

Takeaway: Kelly's decade-old bet that adaptability beats scale in the AI era has held; audit one manual workflow this quarter before agentic tools make the decision for you.