Peter Thiel said in 2011 that the future had stalled. Ray Kurzweil said in 2005 that the future was on schedule. Reading both texts now, with a 9.3-million-patent corpus running underneath them, the answer is that they were arguing about different decades — and that the inflection between those decades happened, with disconcerting precision, the year after Thiel published.

In October 2011, Peter Thiel published The End of the Future in National Review. The essay opens with Revelation 6:5–6 — the third horseman, the black horse, the scales — and closes with the line everyone remembers: the first and the hardest step is to see that we now find ourselves in a desert, and not in an enchanted forest.

Between those two images, the argument is more careful than the polemics it inspired. Thiel was not saying technology had stopped. He was saying that, since roughly 1973, real progress had been confined almost entirely to information technology, and that a half-century of stagnation in everything else — energy, transportation, biotechnology outputs, civil engineering, real wages — had been hidden by financialization, by inflation in education and health care, and by the mistaken assumption that Moore’s Law would carry the whole economy on its back.

Like Alice in the Red Queen’s race, we (and our computers) have been forced to run faster and faster to stay in the same place.

Six years earlier, in The Singularity Is Near (2005), Ray Kurzweil had argued the opposite. The Law of Accelerating Returns was a fact about the universe. Information technology bent every field it touched onto an exponential. By the late 2020s, the destinations Thiel was mourning — cheap energy, cured cancer, near-AI, the reversal of biological aging — would all arrive on schedule. Regulation, in Kurzweil’s framing, was a stone in a stream. The water always went around it.

It is now 2026. We have spent the last six months scoring Kurzweil’s predictions against a database of 9.3 million US patents, 357 million academic papers, and roughly half a million clinical trials. The full breakdown lives on our Kurzweil Scorecard. What that audit suggests, when held up against Thiel’s actual 2011 diagnosis and the Founders Fund manifesto that Bruce Gibney wrote shortly after, is that the two men were not arguing about whether technology was decelerating or accelerating. They were arguing about whether the deceleration was structural or transient.

What Thiel actually said, with the patent record running underneath

Thiel’s load-bearing economic claim is that the single most important economic development in recent times has been the broad stagnation of real wages and incomes since 1973. He pinned that inflection on the oil shock and on a parallel collapse in civilian-sector technological productivity. The supporting evidence list, in his own words:

Travel speed peaked. The centuries-long acceleration of travel speeds … reversed with the decommissioning of the Concorde in 2003. No civilian aircraft has flown supersonic over the United States since.

Energy got more expensive. Real oil prices today exceed those of the Carter catastrophe of 1979–80. Nixon’s 1974 call for full energy independence by 1980 has given way to Obama’s 2011 call for one-third oil independence by 2020.

The cancer war stalled. In 1970, Congress promised victory over cancer in six years’ time; four decades later, we may be 41 years closer, but victory remains elusive and appears much farther away.

Pharma was liquidating its labs. In the next three years, the large pharmaceutical companies will lose approximately one-third of their current revenue stream as patents expire, so, in a perverse yet understandable response, they have begun the wholesale liquidation of the research departments that have borne so little fruit in the last decade and a half.

Computers were the lonely exception. Cellphones in 2011 contain more computing power than the entire Apollo space program in 1969.

The patent record from 1973 to 2011 mostly agrees with him. US aerospace patents fell as a share of total grants for two decades. No new commercial nuclear reactor was ordered between 1978 and 2012. Eroom’s law — Jack Scannell’s 2012 Nature Reviews Drug Discovery paper — showed inflation-adjusted FDA approvals per billion R&D dollars halving roughly every nine years from 1950 to 2010. Median real wages in the US grew about 10 percent between 1973 and 2010, against roughly 0.7 percent per year for mean wages. Thiel’s economics were not contested by the data; they were contested by the consensus.

The most prescient passage in the essay is the one almost no one quoted at the time:

Leverage is not a substitute for scientific progress.

Thiel argued in 2011 that the 1990s tech bubble and the 2000s housing bubble were both symptoms of the same underlying phenomenon: when real returns from technology fall below the expectations of pension funds and other investors, capital reaches for leverage to make up the gap. The 2011–2024 era of zero-rate cheap money, SaaS multiples, private-market valuations, and crypto was exactly the next chapter of that story. He called it.

What Founders Fund specifically bet, and how those bets aged

The Founders Fund manifesto, written by Bruce Gibney, was the more actionable companion piece. It identified four sectors where the manifesto argued the future was unbuilt and the capital was misallocated: aerospace and transportation, biotechnology, advanced machines and software, and energy. Each section made specific, testable claims. Fifteen years later, those claims are unusually scoreable.

Launch costs. One of the major barriers to making use of space is the sheer cost of getting material into orbit: about $19,000 per kilogram … SpaceX appears to be on track to reduce costs by that order of magnitude. This is the cleanest hit on the page. SpaceX Falcon 9 in 2025 prices Starlink-class payloads to LEO around $1,500/kg, with Starship targeting under $200/kg at maturity. The order-of-magnitude reduction the manifesto bet on has been delivered, with another order plausibly in flight. The patent record reflects this: SpaceX alone now holds more granted launch-vehicle and reusable-stage patents than the entire US prime-aerospace block held in 2010.

Genome sequencing. Present methods of sequencing (which use fluorescence) can only sequence about 95% of larger genomes, take forever to do so, and cost a fortune. In 2011, a human genome cost roughly $10,000 and took weeks. By 2014 the $1,000 genome arrived. Ultima Genomics announced a $100 genome in 2022. Oxford Nanopore’s long-read MinION ships for $1,000. The “lack of data” bottleneck the manifesto called out has collapsed by roughly two orders of magnitude on cost and one on accuracy.

The “medieval” drug-discovery critique. Discovery still proceeds by enlightened guesswork, rather than as a disciplined process — and there is no good way for investigators to share data. AlphaFold 2 (2021) collapsed protein structure prediction; AlphaFold 3 (2024) extended to protein-ligand binding. Recursion’s industrialized phenomic-screening platform now runs ~2.2 million experiments per week. Insilico’s INS018_055 was the first wholly AI-designed drug to enter Phase 2 trials (2024). The manifesto’s “medieval guesswork” critique was substantially correct in 2011 and is no longer the right description in 2026.

Robotics. The industry remains over-focused on producing vanity robots with hyper-specific capability — clunky simulacra that play the violin or smile pointlessly — rather than solving more general problems, like locomotion. Half-paid. Boston Dynamics’ Atlas, Figure 02, Tesla Optimus, and Unitree H1 are all general-locomotion humanoid platforms; Waymo crossed 100 million fully autonomous passenger miles in 2024. But commodity-priced general-purpose robots — the manifesto’s deeper test — are not yet here. The patent count for “humanoid bipedal” went from 14 in 2014 to 178 in 2025, which is the right slope on the right axis but not the destination.

AI. The field remains relatively poorly funded, a surprising result given that the development of powerful AIs … would probably be one of the most important and lucrative technological advances in history. The most spectacularly wrong prediction in the manifesto, in the most useful direction. AI funding in 2024 alone was approximately $100 billion globally. The transformer architecture (2017) and the GPT/Claude/Gemini line of models have absorbed the hyperscale-compute capacity of the entire industry. The diagnosis was right; the timeline was off by roughly a decade in the favorable direction.

Energy. We have made little progress in generating more energy more cheaply. The most contested. Levelized cost of utility-scale solar fell about 90 percent between 2010 and 2024. Lithium-ion battery cells fell roughly 90 percent over the same window. But the manifesto’s complaint was specifically that alternative technologies for actually generating energy have not produced particularly good returns — and that critique has held up better than the LCOE numbers suggest, because the rents from cheap solar and cheap batteries have largely accrued to Chinese manufacturing, not to the US clean-tech investors who took the original risk. Thiel’s clean tech has become a euphemism for energy too expensive to afford aged poorly on the engineering side and well on the venture-returns side.

Supersonic flight. Boom Supersonic’s XB-1 demonstrator went supersonic over the Mojave on January 28, 2025. FAA Part 91.817 — the 1973 ban on civil overland supersonic flight that Thiel singled out as a symbol of regulatory ossification — was repealed by executive order on June 6, 2025. Boom Overture is targeting commercial service later this decade. The unbuilt future Thiel mourned is now under construction.

Nuclear. NuScale’s small modular reactor received NRC certification in 2023 (the first SMR ever certified in the US), though its inaugural project was cancelled later that year on cost grounds. Oklo’s combined construction-and-operating license application is in front of the NRC. X-energy’s Dow Seadrift project broke ground in 2025. Real, recent, slow — exactly the pattern you would predict if you took Thiel’s one cannot in good conscience encourage an undergraduate in 2011 to study nuclear engineering as a career line literally and asked when the regulatory thaw would begin.

The aggregate verdict on the Founders Fund prediction list is roughly: launch costs paid, sequencing paid handsomely, drug discovery paid via AI, AI itself paid 10x, robotics paid partially, energy generation paid via solar/batteries but not the way the manifesto framed it, supersonic and nuclear are unblocking now. That is a vastly better track record than the standard ten-year retrospective on a venture manifesto.

Why did the curve bend right after Thiel published?

This is the question that should make a careful reader uncomfortable. The End of the Future came out in October 2011. The Founders Fund manifesto followed shortly after. And then, almost on cue, the curve started bending the other way. Why?

Five candidate explanations, in declining order of how confident the patent record makes us about them:

1. The AlexNet inflection was a coincidence in calendar time but a causal trigger

In December 2012 — fourteen months after Thiel’s essay — Krizhevsky, Sutskever, and Hinton’s convolutional neural net cut the ImageNet error rate roughly in half. That single result was the starting gun for the deep-learning era. The 2017 transformer paper, GPT-3 in 2020, and ChatGPT in late 2022 followed on a roughly five-year doubling cadence. AI is not just one of the Founders Fund bets — it is the input to most of them. AlphaFold made the medieval-guesswork drug-discovery critique obsolete. GNoME turned materials chemistry into a search problem. AI-controlled stabilization made fusion plasma confinement tractable. Bits got cheap and dense enough to start designing atoms because a 2012 vision-network paper started a chain that ended in foundation models. The patent record carries the signal cleanly: US AI-related patent grants went from roughly 4,000 in 2012 to over 60,000 in 2024.

2. SpaceX produced an existence proof, and capital followed

SpaceX’s first ISS resupply mission flew in May 2012, seven months after Thiel’s essay. The first booster recovery was December 2015. Once the order-of-magnitude launch-cost reduction was demonstrated, the calculation for funding hard physical-world startups changed everywhere. This is a feedback loop more than a coincidence — Thiel was the first outside investor in SpaceX, and SpaceX was the first existence proof for the manifesto thesis. Anduril (2017), Varda (2020), Commonwealth Fusion (2018), Helion, Hadrian, Castelion, and Saronic all trace their fundability back to the moment SpaceX showed the model worked.

3. The “it’s time to build” coalition redirected attention and capital

Gibney’s manifesto, Marc Andreessen’s 2020 “It’s Time to Build”, Patrick Collison and Tyler Cowen’s 2019 “We Need a New Science of Progress”, the Stripe Press canon, and a16z’s “American Dynamism” thesis were all downstream of the Thiel/Gibney diagnosis. The argument partly self-fulfilled because the people who read it ran the next decade of capital allocation. Founders Fund itself, Lux, the new generation of hard-tech specialists (Eclipse, 8VC’s industrial book, Lightspeed’s defense practice) collectively pushed several tens of billions of dollars at the exact sectors the manifesto named. You should not over-credit the polemic — but you also should not pretend it had no causal effect when the LPs writing the checks read it.

4. China stopped being a customer and started being a competitor

The post-2010 reality of China as a peer in chips, batteries, EVs, biotech, and now AI forced a Cold-War-style refocus on frontier physical-world technology. The CHIPS Act (2022), the Inflation Reduction Act (2022), the rebuild of the defense industrial base, and the FAA Part 91.817 repeal (2025) were all responses to a competitive pressure that did not exist in 2011. The Manhattan Project comparison Thiel made in his closing section turned out to be the right historical analog — but the catalyst was Beijing, not Washington.

5. Independent cost curves crossed during the same window

Some inflections would have happened regardless of any essay. The $1,000 genome arrived in 2014 because Illumina’s HiSeq X iterated on its own physics. Lithium-ion battery cell costs fell 90 percent because of Chinese manufacturing scale, not because of US venture allocation. Solar PV fell along the same curve. These are the cases where Kurzweil’s law-of-accelerating-returns is the better description of what happened.

What none of the five explanations supports is the lazy reading that the inflection was inevitable. It was not. Most of the people who read The End of the Future in 2011 dismissed it as Tea Party-adjacent doom-mongering. The Founders Fund manifesto was treated as a polemic, not a prediction. The fact that the prediction list paid out so much better than the dismissal list is the part of the story that should haunt the consensus.

The uncomfortable corollary is that the next inflection — the one that turns the Thielian-blocked sectors (housing, civil nuclear at scale, transit construction costs, organ markets) into Kurzweilian acceleration — is probably waiting on the same five ingredients in some new combination. A trigger technology (AI-on-physics is the obvious candidate). An existence proof (a single American city that actually builds; a single SMR that ships at cost). A capital redirection (the abundance/build coalition is the candidate). A geopolitical forcing function (already here). And independent cost curves crossing under the noise floor. Watch for those five. They are the same five that bent 2012.

What the Kurzweil Scorecard finds, fifteen years after Thiel wrote

The Scorecard’s central pattern is consistent across the eleven scorecards published so far. Kurzweil’s targets — the destinations he named — were arriving roughly on his predicted timeline. His mechanisms — the specific platforms and companies he identified — were almost always wrong.

A non-exhaustive list of the targets he named in 2005 and where they actually landed:

Human-level performance on language and reasoning by the mid-2020s. Arrived. Mechanism: transformer language models, not the genetic-algorithm-evolved neural nets he sketched.

A drug that targets the molecular root of Alzheimer’s. Arrived July 2023. Lecanemab (Leqembi) binds the soluble amyloid-beta protofibrils Kurzweil singled out by name. Mechanism: monoclonal antibody, not the proteasome-repair pathway he emphasized.

Individualized brain stimulation that durably treats depression. Arrived September 2022. Stanford’s SAINT protocol, FDA-cleared as Magnus Medical’s offering, achieved 90.5% remission in treatment-resistant depression in the pivotal trial. The “virtual lesioning” Kurzweil described in 2005 is now a billing code.

Reprogramming somatic cells back to pluripotency. Arrived August 2006, when Yamanaka and Takahashi published the Oct3/4-Klf4-c-Myc-Sox2 cocktail. The 2006 Cell paper has 26,162 citations by OpenAlex’s count. Kurzweil predicted in 2005 that the Bush-era embryonic-stem-cell restrictions would push the field toward exactly this kind of transdifferentiation. He was right within twelve months.

In-bloodstream cancer killing on something like a switch. Arrived March 2024 in patent form (US 11,931,465: Prussian blue nanoparticles + checkpoint inhibitor + near-infrared laser). Nanospectra’s gold-silica nanoshell prostate-cancer trial reported 73% biopsy-confirmed disease elimination at 12 months in 2024. Mechanism: a nanoparticle plus an off-the-shelf antibody, not a Drexlerian assembler.

Atmospheric water harvesting at meaningful scale. Arrived 2025. MOF-303 and MOF-801 chemistry, GE Vernova joint venture, Texas data-center contracts. Yaghi’s 2025 Nobel formalized a twenty-two-year arc from a curiosity paper to a working unit on a test pad.

Solar at a price that beats fossil for new build. Arrived. Perovskite-silicon tandem cells are the next layer.

The single most striking line in The Singularity Is Near is also the one Thiel implicitly bet against: Despite stem-cell restrictions, cell-therapy research and the broader biotechnology field have not been affected to a significant degree … restrictions on embryonic stem-cell research have accelerated alternative approaches such as transdifferentiation. In print, in 2005, twelve months before Yamanaka. The patent record now shows iPSC papers going from 65 per year in 2005 to 2,884 in 2025, while embryonic-stem-cell publications collapsed. (Full evidence on the regulation scorecard.)

The honest read is that Thiel was right about the 1973–2011 era and Kurzweil’s destinations were right about the 2012–2030 era. The bridge between them — and this is the part both books undertold — was that the bits eventually got cheap and dense enough to start designing the atoms. AlphaFold collapsed protein structure prediction in 2021. DeepMind’s GNoME identified 2.2 million new stable crystal candidates in 2023, expanding the known materials space by an order of magnitude in a single paper. The MOF chemistry that won Yaghi the Nobel is being optimized today by combinatorial-screening pipelines that would have taken a postdoc career in 2005. The mechanism Kurzweil named for this transition was nanobots. The mechanism that arrived was GPUs.

The free-market question, on the data

The original prompt for this post was the natural free-market reading of Thiel: how much of what looks like stagnation is actually that the right ideas are sitting in the literature, but we have made the commercial path illegal — or so expensive through clinical trials, zoning, and licensing that capital walks away?

The patent record gives a more nuanced answer than either Thiel or Kurzweil gave.

Where the regulation-suppresses-the-frontier story is clearly true

Civil supersonic flight was illegal in US airspace from 1973 until June 2025; the patents during that window are mostly military. Commercial nuclear faced a de facto licensing freeze for thirty-five years. Organ markets are illegal under the National Organ Transplant Act, which is one reason there are about 100,000 Americans on the kidney waiting list and roughly six thousand of them die each year. Civil drone airspace integration ran a decade behind the underlying technology because Part 107 had to be written. Housing in coastal cities is the cleanest case: zoning is the binding constraint, not patents and not engineering. In each of these, you can read a thirty-year frontier of unbuilt patents and ask what would have happened if the rules had been different. Thiel’s instinct here is correct, and the recent unblocks (XB-1 supersonic, NuScale SMR cert, Oklo, drone Part 108) are exactly what a free-market reading would predict the moment the friction lifts.

Where the regulation-as-stone-in-stream story is clearly true

Embryonic stem cell restrictions in 2001 led directly to the iPSC discovery in 2006. The CRISPR somatic-vs-germline split is similar: the 2019–2025 international moratorium on germline editing held, but somatic CRISPR raced through to FDA approval (Casgevy, 2023). California’s Proposition 71 in 2004 authorized $3 billion in state stem-cell funding precisely to route around the federal restrictions; CIRM has now funded 113 clinical trials and seeded fifty-plus startups. When the underlying biology is not banned outright, capital and talent reroute around the regulation faster than the regulator can rewrite it. Kurzweil’s instinct here is correct.

Where the picture is mixed

The EU AI Act, in force since August 2024, did not stop frontier AI; it exported it. Europe hosts none of the world’s ten largest AI companies. In 2024, the US attracted roughly $29 billion in AI investment versus $1.5 billion for Europe — about a 20x gap. By November 2025 the EU was already moving to weaken the law. The technology accelerated; the rents went somewhere else. The US chip export controls aimed at China produced a similar pattern in reverse: Huawei stockpiled HBM ahead of the cutoff, and DeepSeek shipped a frontier-class model trained on chips that were only marginally legal to acquire. Regulation slowed the laggard; it did not stop it; it midwifed an indigenous Huawei/DeepSeek stack that is now a credible competitor to Nvidia.

The integrated model

If you put Thiel’s 2011 diagnosis on one axis and Kurzweil’s law-of-accelerating-returns on the other, the past twenty years look less like a debate and more like two halves of one curve.

From roughly 1973 to roughly 2012, Thiel’s diagnosis was the better description: bits raced, atoms stalled, regulation accumulated, productivity in physical-world R&D fell, and the gap was papered over with leverage. The patent data shows it. The drug approval data shows it. The transportation data shows it. The wage data shows it. Leverage is not a substitute for scientific progress is the line that ages best.

From roughly 2012 forward — and accelerating sharply from about 2020 — the Founders Fund manifesto’s bets started paying out, sector by sector. SpaceX delivered the order-of-magnitude launch-cost reduction. Genome sequencing collapsed by two orders of magnitude. AlphaFold answered the medieval-guesswork critique. Boom and NuScale and Oklo are now trying to convert the regulatory thaws into actual hardware. Robotics is half-built. The Kurzweil destinations — lecanemab, SAINT, iPSC, CAR-T, GLP-1 — are arriving on schedule, almost always wearing a different mechanism than the one Kurzweil named.

The unfinished business is that the regulatory layer compounded during the years bits were running and atoms were not. By the time bits came for atoms, atoms were sitting under thirty extra years of NRC, FDA, NEPA, zoning, NOTA, and Part 91.817. The industries best positioned to absorb the AI-on-atoms wave are now the ones with the lightest regulatory overhang (compute, software, materials chemistry done in private labs) or the ones where capital is large enough to amortize the friction (large-cap pharma, hyperscale cloud, EVs, defense primes). The industries worst positioned are the ones where the regulation is binding and coordinated and attached to physical siting (civil aerospace, civil nuclear, transit, urban housing). That last bucket is exactly the bucket Thiel kept pointing at — and the leading indicator that it is opening again is the small set of recent regulatory unblocks named above.

We are still in the desert Thiel described. But the satellite imagery is starting to show green at the edge.

Method

Patent counts come from our mirror of the USPTO utility grant feed (~9.3 million documents through 2025). Citation counts come from our OpenAlex snapshot (~357 million scholarly works). Clinical trial counts come from the full ClinicalTrials.gov dataset (~541,000 studies). Specific patent numbers (US 11,931,465; US 11,111,542; US 12,178,885; US 10,683,644; US 12,372,254; US 12,115,383; US 8,058,065) and clinical-trial identifiers were pulled and read directly. The Kurzweil prediction set and verdicts come from our internal singularity-tracker batches; per-batch detail and full evidence chains are linked from the Kurzweil Scorecard. Quotations from Peter Thiel, The End of the Future (National Review, October 3, 2011) and from the Founders Fund manifesto authored by Bruce Gibney are taken from the published texts. Eroom’s law is from Scannell et al., Nature Reviews Drug Discovery 11 (2012). Sequencing cost benchmarks are from NHGRI and from each company’s product announcements. Launch-cost figures are from SpaceX public materials and FAA filings. EU AI Act, chip export, and CRISPR moratorium figures are documented in the Scorecard’s regulation post. This is a description of what has been filed, published, approved, ruled on, and flown — not investment advice.