How fast is AI? A living map of the AI change curve<br>Explore the curve<br>How fast is AI?<br>You are here on the AI change curve.<br>For 150 years, US living standards rose ~2% a year, through electricity, the transistor, and the internet. AI may be the next of them. The question isn't whether it's powerful, but how fast that power turns into growth. The answer hides in a single idea: weak links .<br>Slow, real upside. Fast, fragile downside. Does AI finally bend the 2% line, or just hold it? Here's the evidence, the forecast, and the hardest objections, updated as the news arrives.

US income / person (history)World income / person≈2% / year trendBaselineFull automationIncomplete automation

Source: US and world history approximated from Maddison/BEA patterns; the world line grows slower early, then accelerates as the rest of the world catches up. Forward paths follow the published “Continuing the Past” growth scenarios. · Aghion, Jones & Jones (2019)

~100M<br>ChatGPT users<br>in ~2 months, fastest ever at the time

150 yrs<br>of ~2% growth<br>through every prior transformative tech

4.3% → 3.0%<br>computer share of GDP<br>falling since 2000, despite more chips

20+ yrs<br>self-driving diffusion<br>from “solved in 5” to still-rare

The big idea<br>A chain is only as strong as its weakest link.<br>Business success means completing many tasks in a row. Make most of them incredibly strong and the chain barely improves: the few weak links still set its strength. Pick a business below, automate the tasks AI is already good at, and watch the overall strength refuse to budge.<br>Pick a businessSoftware companyHospitalBankRetailer<br>To ship a product feature , this business runs 9 tasks in a chain. AI makes the routine ones strong; the human judgment, accountability, and physical tasks stay weak.

38<br>Specs

58<br>Design

95<br>Code

82<br>Review

88<br>Tests

92<br>Deploy

85<br>Monitor

35<br>Roadmap

42<br>Customer

Each bar is a real task. Click to automate it to 100%.

Overall chain strength<br>59%

0 of 9 tasks automated

Getting stronger. Keep raising the lowest tasks, not the ones that are already high.<br>Automate the strong tasksFix the weak links tooReset

A billion times the compute<br>The compute a dollar buys has doubled every couple of years for decades, now roughly a billion times what the mainframe era got. Yet productivity growth still sits near 2%. The abundant input gets cheap; judgment, attention, and trust, the weak links, still gate the result. More FLOPs don't buy more wisdom.

Free doesn't mean infinite<br>A clean result hides in the model: drive one task's cost to zero and you raise total output by only that task's share of the economy. If software is a few percent of GDP, infinitely cheap software makes us a few percent richer, once. Real growth means automating the next weak link, and the next.

The puzzle<br>Transformative technologies, and still 2% a year.<br>Electricity, internal combustion, antibiotics, semiconductors, the internet: each wildly transformative, yet growth never strayed far from 2%. Each new technology kept 2% going for another 50 years as the previous one ran out of steam.

Real income / person ≈2% / year<br>ErasElectricity1882Mass production1908Transistor1947Microprocessor1971The Internet1991Generative AI2022

Source: US real income per person, 2025 dollars (approximated for display).

Computers are everywhere, but their price falls faster than their quantity rises. The weak links capture the value.

Source: Computers’ share of US GDP, approximated from the BEA/BLS value-added pattern.

The chart that flips the worry<br>Computers are the most plentiful thing in the economy, and their share of GDP has fallen by a third since 2000. Price falls faster than quantity rises. The plentiful thing gets cheap; the scarce thing, humans and the weak links they hold, captures the value.<br>When you worry AI will automate everything, remember: abundance drives down the price of what's abundant. Scarcity is where the returns go.

Where we are today<br>The concrete record: milestones, predictions, bottlenecks.<br>Not abstractions. Real dates, real numbers, and a scoreboard of who called it right. The pattern: adoption can be lightning-fast while economic transformation stays slow.<br>Milestone timeline<br>Mar 2004Verified<br>DARPA Grand Challenge: zero finishers<br>Not a single autonomous vehicle completed the desert course. The starting gun for self-driving, and a reminder of how hard the physical world is.<br>DARPA Grand Challenge (2004)

Oct 2005Verified<br>Stanford's "Stanley" wins the DARPA Grand Challenge<br>One year after zero finishers, Sebastian Thrun's team completed the 132-mile course. Twenty years later, robotaxis are still rare outside a few cities.<br>DARPA Grand Challenge (2005)

Sep 2012Verified<br>AlexNet ignites the deep-learning era<br>A deep neural network crushed the ImageNet benchmark, kicking off the modern wave of AI capability gains.<br>ImageNet / AlexNet

Mar 2016Verified<br>AlphaGo defeats Lee Sedol<br>DeepMind's system beat a top human Go player 4–1, years ahead of expert expectations for the...