Steven Pinker’s Enlightenment Now proves that, on a 200-year baseline, almost everything is up and to the right. Peter Thiel’s The End of the Future argues that, since 1973, almost nothing in the physical world is. Both can be true. The trick is to plot each metric on its own longest available baseline and see which regime you’re in for that specific axis. Here are nineteen of the cleanest charts the data allows, with a verdict on each.

The previous post in this series argued that Thiel’s 2011 diagnosis was substantially correct as a description of 1973-2011, that the Founders Fund manifesto’s specific bets paid out, and that the inflection back toward the Kurzweil exponential happened, with disconcerting precision, the year after Thiel published. (Full piece here.) That argument needed a backbone. This is the backbone.

Each chart below is annotated with two shaded regions: a grey band for what we are calling the Thielian plateau (1973–2012), and a purple band for the post-2012 resumption window. The pattern that emerges across nineteen metrics is more interesting than either Pinker or Thiel told you on their own.

Tier 1 — The Pinkerian baseline

Start with the metrics where the long-baseline rise is so dominant that the 1973 inflection is barely visible. These are the charts that vindicate Enlightenment Now — and you have to put them first, because any honest stagnation analysis has to acknowledge the floor it stands on.

Real US output per person is up roughly seventy-fold since 1700 and sevenfold since 1900. On a log scale the line is almost straight. The 1973 oil shock is visible as a slight bend, not a break. If you only ever saw this chart, you would dismiss the entire Thiel argument as cope. Most of Pinker’s book is built on charts like this one.

The single most important chart in human history. Doubles between 1800 and 2010. The most recent points are the only honest blemish: the COVID dip in 2021 (76.4 years) was the largest peacetime drop on record, and the 2023 recovery to 78.4 years is still below the 2019 peak. But the 200-year shape is unambiguous.

William Nordhaus’s famous chart, extended through the LED era. Hours of work to buy 1,000 lumen-hours of light fell roughly 100,000-fold between 1800 and 2024. Light is the cleanest possible illustration of why GDP statistics undercount progress — the price index thinks “a candle” and “a 100-lumen LED” are the same product. The post-2012 LED transition is another order of magnitude on top of an already vertiginous curve.

The lonely exponential Thiel conceded in his 2011 essay. It is also the input to almost every post-2012 resumption chart further down. Note the slight upward bend post-2012 — that is the GPU/specialized-silicon era pulling the curve away from the historical Moore’s Law slope of pure CMOS scaling. Computing per dollar is now improving roughly twice as fast as it was in 2010.

Tier 2 — Where Thiel’s 1973 inflection shows up cleanly

These are the charts that anchored Thiel’s argument. The kink at 1973 is real and visible. None of them have recovered the pre-1973 slope, even with the post-2012 resumption.

This is the chart Thiel built his economics on. From 1947 to 1973, real US median household income roughly doubled. From 1973 to 2010, it grew by about 9 percent — over thirty-seven years. The post-2012 window has produced a real gain (from $65,500 in 2010 to $80,610 in 2023), and the curve is now back above its 1999 peak in real terms. But the 1947–73 slope has not returned.

Thiel’s load-bearing energy chart. Oil sat around $20–30/bbl in real terms for over a century before 1973. Since then it has spiked four times above $100, the mean price has roughly doubled, and the variance has exploded. Shale unlocked some supply, but the regime change is permanent. This chart is also the cleanest test of one of Thiel’s strongest claims: real oil prices today exceed those of the Carter catastrophe of 1979–80. In 2008 and 2014, that was true. In 2024, with US shale at full output, it is no longer true. Mixed verdict.

The chart Pinker doesn’t show. Life expectancy at birth is a triumph; life expectancy at 65 is a much harder slog. The big gains in the headline number came from eliminating childhood mortality, not from extending old age. The post-1980 slope is real but slow — about one extra year of life per 15 years of progress — and the post-COVID 2023 reading is actually below the 2019 peak. This is closer to the chart Thiel had in mind when he wrote that the cruder measure of U.S. life expectancy continues to rise, but with some deceleration.

Tier 3 — The pure regulatory plateau

These charts are not about whether the technology works. They are about whether we let it.

The single most striking chart in Thiel’s essay, redrawn. From 1800 to 1969, the maximum civilian travel speed compounded for 169 straight years — sail, rail, prop, jet, supersonic. The Concorde topped out at Mach 2.04 and entered commercial service in 1976. It was retired in 2003. From 2003 to 2025, the maximum civilian travel speed available to a paying passenger has been the same as the cruise speed of a Boeing 747-400 designed in the 1980s. Boom Supersonic’s XB-1 went supersonic over the Mojave on January 28, 2025; FAA Part 91.817 was repealed in June 2025. The next data point is plausibly 2029.

The cleanest chart in the deck. The number of commercial nuclear reactors operating in the United States peaked in 1990 and has been falling slowly ever since. There has been one new reactor (Vogtle 4) added to the grid since the 1990s. Three Mile Island happened in 1979 — but the licensing freeze it triggered is, by 2025, longer than the entire history of commercial nuclear power up to that point. NuScale’s NRC certification in 2023 is the lead indicator that the regime is changing. The data point that would matter is the next number on this axis going up.

The chart that gets worse. US subway and heavy-rail construction cost per mile, in real terms, has risen roughly twentyfold since the early 20th century. New York’s Second Avenue Subway Phase 1 (opened 2017) ran $2.6 billion per mile, against $50–80 million per mile for the original IRT and BMT lines a century earlier. This is not a technology story — Spain and South Korea build subways for $200–400 million per mile today. It is a regulatory, procurement, and political story. The post-2012 resumption window is invisible here because it has not yet happened.

Tier 4 — The Eroom puzzle

Eroom’s law — Moore’s law spelled backwards. From 1950 to 2010, the number of new molecular entities approved by the FDA per inflation-adjusted billion dollars of pharma R&D fell by a factor of about forty. Scannell’s 2012 paper made this chart famous; Thiel cited it implicitly. The interesting question is whether the curve is finally inflecting. The 2024 data point ticks up for the first time in seventy years. Cell and gene therapy approvals (CAR-T platforms, Casgevy, Zolgensma, Luxturna), the GLP-1 weight-loss class, and the first AI-discovered drugs in clinical trials are the candidate explanations. It is too early to call this an inflection. It is not too early to call it a candidate inflection.

Tier 5 — The post-2012 resumption

Now the charts the Founders Fund manifesto bet on. Each one shows a clean inflection inside the purple band.

The cleanest “manifesto bet paid” chart on the page. From 1957 to about 2010, the cost to put one kilogram into low Earth orbit was approximately constant at $10,000–$20,000. SpaceX’s Falcon 9 (2010) cut that in half. Reusable Falcon 9 (2018) cut it by another factor of four. Falcon Heavy and dedicated Starlink missions have driven the marginal cost lower still. Starship, when it reaches operational maturity, targets under $200 per kilogram — an order of magnitude beyond where the curve sits today. This is what a regime change looks like in chart form.

The fastest cost-down curve in the history of any technology. The first human genome cost roughly $3 billion (the Human Genome Project, 1990–2003). The $1,000 genome arrived in 2014. Ultima Genomics announced a $100 genome in 2022. This curve beat Moore’s Law by a factor of about 10x during the 2007–2014 sequencing-platform transition (Solexa → Illumina HiSeq → NovaSeq). The Founders Fund manifesto in 2011 specifically called out the slow, expensive, inaccurate state of sequencing as a bottleneck. By 2014, that bottleneck was gone.

Solar PV modules at $0.10/W in 2024 are roughly 800x cheaper than they were in 1976. The curve is the cleanest documented example of a learning rate (about 23% cost reduction per doubling of cumulative capacity) operating uninterrupted for almost five decades. The Founders Fund manifesto was skeptical that incremental solar work would beat fossil; the curve says it did, partly because the rents went to Chinese manufacturing rather than US venture portfolios.

Lithium-ion battery pack costs fell from approximately $7,500/kWh at first commercial introduction (Sony, 1991) to $115/kWh in 2024. This is the single chart that explains the entire EV transition, the buildout of grid-scale storage, and the economics of solar+storage as a baseload-replacement combination. The post-2012 inflection is sharp because that is when EV manufacturing scale began to dominate the supply curve.

Frontier AI training compute has doubled roughly every six months since 2012 — twice the rate of historical Moore’s Law, sustained for over a decade. AlexNet (2012) used about 10¹⁵ FLOP. GPT-4 (2023) used about 10²⁵. The 2024 frontier crossed 10²⁶. This is the curve that turns Kurzweil’s Law of Accelerating Returns from a metaphor into an engineering specification. It is also the input to most of the other resumption charts on this page — AlphaFold, GNoME, AI-discovered drugs, AI-controlled fusion, AI-designed catalysts.

The chart Thiel called out by name when he cited the 1971 War on Cancer. Five-year relative survival across all cancer sites combined rose from 49% in 1975 to about 69% in 2020. That is a real, hard-won 20-percentage-point gain — but it took fifty years and most of it predates the Thiel essay. The post-2012 slope flattens, partly because the easy gains from screening (cervical, colorectal) have already been captured and the new gains have to come from the hardest cancers (pancreatic, glioblastoma). Lecanemab, CAR-T, and the immune-checkpoint era are pushing on the curve, but the chart will not show that for another decade.

Tier 6 — The honest mirror

Two charts that make any optimistic resumption story uncomfortable. Neither has inflected. Both are arguably the most important metrics on this page.

Total factor productivity is the residual — what is left of GDP growth after you account for labor and capital inputs. It is the closest thing economists have to a measure of “are we actually getting better at making things.” The 1947–73 golden age ran at about 2.3% per year. The post-1973 average has been roughly 0.8%. The brief 1995–2005 IT-investment boom is the only sustained recovery, and it ended. As of 2024, the 10-year rolling number is back above 1% — but well below the 1947–73 baseline. This is the chart that should haunt anyone who thinks the post-2012 resumption has fully arrived.

The most civilizationally consequential chart on this page. The US total fertility rate has been below replacement (2.1) for almost every year since 1972. The 2024 reading of 1.62 is the lowest in recorded US history. South Korea is at 0.72. China is at 1.0. This is not a technology problem in the conventional sense — but it is exactly the kind of metric Pinker’s narrative does not engage with, and exactly the kind of metric a serious stagnation analysis has to put on the table. The post-2012 resumption is invisible here. The trend line is going the wrong way.

The pattern across nineteen charts

If you stand back from the individual metrics and look at the verdict tags, three patterns emerge.

The further upstream the metric, the more Pinkerian it looks. The basic inputs to civilization — calories per dollar, lumens per dollar, FLOPS per dollar, real GDP per capita, life expectancy at birth — are all up and to the right on a 200-year baseline. None of them shows a meaningful 1973 break. This is the foundation Pinker built his book on, and the foundation is sound.

The further downstream the metric, the more Thielian it looks. The metrics that aggregate civilizational performance into a single number — total factor productivity, real median wages, life expectancy at 65, total fertility rate — all show a clean break around 1973 and have not recovered the pre-1973 slope. These are exactly the charts you would expect to see if Thiel were right that we have been substituting leverage and financialization for real productivity gains.

The post-2012 resumption is real but uneven. Five charts show a sharp inflection inside the purple band: launch costs, genome sequencing, solar, batteries, and AI training compute. Three more (Eroom, life expectancy at 65, real wages) show a possible inflection that needs another decade of data to confirm. Three (nuclear reactors, transit construction costs, fertility) show no inflection at all — and these are exactly the metrics most exposed to the regulatory layer Thiel pointed at.

The synthesis from the previous post in this series holds: Thiel was right about the era he wrote in. Kurzweil’s destinations are arriving in the era we just entered. The regulatory layer is the binding constraint where it is binding-and-coordinated-and-physically-sited, and a routable nuisance where it is not. The bridge between the two eras was AI getting cheap enough to design atoms — and that bridge is visible in five of the nineteen charts above and absent in three of them.

If you are an investor or a policymaker, the small-multiples grid is the operational tool. Find the chart that matches your sector. Look at where the inflection is or is not. If your sector is on the upswing inside the purple band, you are probably running too cautiously. If your sector is still flat in the grey band, your job for the next ten years is to figure out which of the five 2012-trigger ingredients (a trigger technology, an existence proof, capital redirection, geopolitical pressure, an independent cost curve) is missing — and to supply it.

We are still in the desert Thiel described. The satellite imagery is starting to show green at the edge. The pattern of the green tells you where to plant.

Method

Each chart is plotted on the longest baseline for which a defensible homogeneous source exists. Where data series cross methodology breaks (e.g., pre- and post-1933 GDP, pre- and post-1947 BLS productivity), we use the standard chained reconstruction. Specific values are rounded to chart-readable precision; for full-precision values the original sources are linked or named in each chart’s source line. The shaded grey region (1973–2012) marks what we are calling the Thielian plateau; the shaded purple region (2012–present) marks the post-2012 resumption window. Verdict tags are our own. Sources include the Maddison Project Database, NHGRI cost-per-genome series, the BP Statistical Review of World Energy, BloombergNEF, IRENA, IAEA Power Reactor Information System, US Census Historical Income Tables, BLS Multifactor Productivity, CDC NCHS Vital Statistics, NCI SEER, Scannell et al. Nature Reviews Drug Discovery 11 (2012), Sevilla et al. (Epoch AI), Nordhaus (1996), and contemporary press for the most recent data points (post-2024). This post is a description of what has been measured — not investment advice.