Kurzweil Scorecard: The Ten Bricks Were Real. The Building Never Came.

In 2003, Nobel laureate Richard Smalley told Eric Drexler that atomically
precise molecular assemblers could not work. Drexler fired back. By 2005,
Ray Kurzweil had picked a side. The final chapter of The Singularity Is
Near is titled “Response to Critics,” and ten of its most concrete
claims form a feasibility argument for Drexler’s vision: small lab
demonstrations that, stacked together, were supposed to show the
Drexlerian nanofactory was only a matter of engineering.

Twenty years later, nine of those ten demonstrations turned out to be real.
The thing they were supposed to enable — a universal, diamondoid-based
molecular assembler capable of building anything atom by atom — did not.
Meanwhile, almost every outcome that assembler was supposed to deliver
(cheap solar, cancer-killing immune machines, pocket supercomputers,
virus-neutralizing drugs, the return of extinct species) either arrived
or is arriving. Just not through Drexler’s mechanism.

The bricks were real. The building he sketched around them never came.

The predictions

Kurzweil’s defense of Drexler rests on specific laboratory results from
1999 to 2004. The logic: if you can rotate a molecule in one direction
with light, abstract a single hydrogen atom with a scanning tunneling
microscope tip, manipulate an atom vertically with a near-contact atomic
force microscope, and model a self-replicating kinematic machine, then
Smalley’s “fat fingers” and “sticky fingers” objections don’t apply and
the broader program should proceed.

In The Singularity Is Nearer (2024), Kurzweil doubles down: “Looking
back almost two decades later, I am pleased to say that recent advances
in nanotechnology are making the ‘top-down’ view look more and more
plausible — even though it will be at least a decade before the field
starts maturing, likely aided by advances in AI.” He moves the timeline
for mature assemblers into the 2030s.

The 2005 micro-claims are testable now. The 2030s forecast is not.

The demonstrations: mostly real, mostly consequential

Kurzweil cited T. R. Kelly’s 1999 Nature paper as the first molecular
system with “unidirectional rotary motion” (ch. “Response to Critics”)
and N. Koumura’s 1999 Nature paper on a “light-driven monodirectional
molecular rotor”. Both are real and both turned out to be seminal. In
2016, Ben Feringa — whose group authored the Koumura paper — shared the
Nobel Prize in Chemistry with Jean-Pierre Sauvage and Fraser Stoddart
for the design and synthesis of molecular machines.

Second-generation Feringa motors are now engineered for applications
Kurzweil couldn’t name in 2005. An April 2024 paper from the Groningen
group added an aldehyde functional group that dramatically improved
photon efficiency. A January 2025 paper coupled rotor rotation to
helicene inversion for unprecedented chirality control. Light-driven
Feringa motors are being explored as mechanotherapeutics — tiny drills
that puncture cancer cell membranes when activated with light.

The Lauhon and Ho 2000 single-hydrogen abstraction via scanning
tunneling microscope holds up. So does Oyabu’s 2003 Physical Review
Letters paper “Mechanical Vertical Manipulation of Selected Single
Atoms by Soft Nanoindentation” (234 citations in our holdings). Carlo
Montemagno’s 1999–2000 biomolecular-motor nanodevices are documented. The
Tihamer Toth-Fejel study on self-replicating nanomachines in kinematic
cellular automata was published as cited.

Freitas’s theoretical arguments — that 0.07 nm positional uncertainty is
only five percent of a 0.3 nm atomic electron cloud, and that Brownian
motion contributes only 0.7 microns of per-second drift for 1-micron
medical nanorobots — remain on the record. Theoretical, not refuted, not
demonstrated at scale.

The building: missing

Kurzweil wrote that by 2005, “theoretical and experimental work had
provided support for positional diamond mechanosynthesis tools including
hydrogen abstraction and carbon dimer placement”. Three years later, the
U.K. Engineering and Physical Sciences Research Council awarded Philip
Moriarty of the University of Nottingham a five-year, £1.53 million
grant specifically to test the Freitas-Merkle minimal toolset
experimentally. This was the experiment.

The diamond portion was abandoned within ten months. Moriarty’s group
shifted to silicon surface work with a qPlus AFM and reported real
progress on silicon dimer manipulation — but the specifically
diamond-based path Kurzweil cited did not yield. Two decades of
literature since 2005 contains 22 papers indexed under “diamond
mechanosynthesis” in our holdings, almost all theoretical, with no
experimental breakthroughs after 2013. Patents referencing diamond
mechanosynthesis since 2005: three.

The one class of “molecular robot” that has scaled is not Drexler’s
diamondoid assembler. It is David Leigh’s chemistry-based molecular
robot, whose 2024 Chem paper describes the Chemputer — a robotic
platform integrating NMR feedback with the XDL chemical programming
language to autonomously synthesize [2]rotaxanes. A programmable
organic-chemistry machine, not a universal atom-placer. It is excellent.
It is not what Kurzweil was defending.

The goals arrived through a different door

The payoffs Drexler and Peterson promised are mostly arriving. They are
arriving through biology, silicon, and lithography — not the diamondoid
nanofactory.

Cheap solar cells as tough as asphalt: utility-scale solar dropped past
$1 per watt and residential is $2.56 per watt in the U.S. in 2025, per
the NREL Spring 2025 Solar Industry Update. The mechanism is silicon PV
with perovskite tandems emerging — perovskite solar cell patents in our
holdings went from one in 2014 to 58 in 2025. No diamondoid
nanoengineering involved.

Molecular mechanisms that kill viruses, biodegrade in six hours: mRNA
vaccines and monoclonal antibodies. Lipid nanoparticles delivering
instructions to the cell’s own ribosomes — a biology-first path Smalley
argued for and Drexler dismissed.

Immune machines that destroy malignant cells on command: CAR-T
therapies are standard of care for several hematologic malignancies. The
top-cited CAR-T paper in our holdings is the 2018 ASTCT consensus
grading paper on cytokine release syndrome (3,149 citations) — a paper
that exists because the technology is established enough to need
toxicity grading. Bispecific T-cell engagers and antibody-drug conjugates
round out the portfolio. None are diamondoid.

Pocket supercomputers: a 2025 smartphone contains roughly 100 billion
transistors fabricated with extreme ultraviolet lithography at 3 nm.
Ahead of schedule — built from silicon CMOS.

Restoration of lost species: Colossal Biosciences, valued at $10.2
billion after a January 2025 Series C, produced gene-edited “woolly
mice” with mammoth-like hair in March 2025, announced a dire wolf
project in April 2025, a moa collaboration with Peter Jackson in July
2025, and an Australian thylacine expansion in August 2025. The tool is
CRISPR and the ancient-DNA pipeline, not nanomachines rebuilding
organisms from carbon feedstock.

End of fossil-fuel use: not yet. Global oil demand hit record highs in
2024. Behind.

The scorecard

What Kurzweil missed, and what he nailed

He nailed the feasibility demonstrations. Every 1999–2003 lab result he
cited to counter Smalley was correctly described. The molecular-machines
program he was defending produced a Nobel eleven years later. STM-based
bond manipulation and near-contact AFM atom placement are routine. The
Freitas and Drexler theoretical arguments on thermal noise and Brownian
motion have not been overturned.

What he missed is how much work a feasibility demonstration does. A
single hydrogen abstraction is not a factory. A single rotor is not a
self-replicating machine. Between a hydrogen-abstraction tip and a
desktop nanofactory sits the problem of generalization, feedstock,
error correction, and scale-up — what Smalley was actually pointing at
with “fat fingers.” The Moriarty program was the honest test. It ran
aground on diamond within a year and had to pivot to silicon.

The uncomfortable lesson for techno-optimist forecasting: the goals
Drexler named are arriving on or ahead of schedule. But the causal
pathway — diamondoid assembler → universal nanofactory → everything — is
not the pathway that delivered them. Biology, silicon, and lithography
got there first. That does not prove Drexler wrong about 2050. It does
say the century-ending view, where a single foundational invention
cascades into everything, keeps failing to predict which foundation wins.

In The Singularity Is Nearer, Kurzweil still writes as if the
assembler is the pathway. If that bet plays out in the 2030s, it will be
a spectacular vindication. If it does not, the 2020s ledger is that
biology and photovoltaics delivered the goal list without ever needing
the building Kurzweil was defending.

Method note

Evidence came from three sources. An internal corpus of roughly 9.3
million U.S. patents and 357 million OpenAlex papers, queried by
keyword and filtered by citation count. A 1,123-prediction extract from
The Singularity Is Near (2005) and the full text of The Singularity
Is Nearer (2024), used to anchor every verdict in Kurzweil’s own
wording. Web research on the Feringa lab, the Moriarty EPSRC project,
Colossal Biosciences, the Leigh group, and 2025 solar cost benchmarks.
All numeric claims were verified against one of these sources this
session.