Kurzweil Scorecard: The Respirocyte Showed Up as a Lipid Vesicle, the Nanobot as an Insulin Pump

In 2025, the US Department of Defense handed a University of Maryland team a $46M contract for artificial blood. The product — ErythroMer — is a hemoglobin-loaded peptide-lipid nanoparticle that stores for years at room temperature and releases oxygen more efficiently than a red blood cell. That same year, Haima Therapeutics collected $4M in DARPA contracts for SynthoPlate, a peptide-coated nanoparticle that mimics platelet adhesion at a bleeding site. Neither product contains a diamondoid gear. Neither is, in Kurzweil’s sense, a nanobot. Both do exactly what he said nanobots would do.

The predictions

Batch 28 of The Singularity Is Near is the “Bridge Three” body-upgrade chapter — respirocytes, vasculoids, artificial platelets, metabolic and digestive nanobots, elimination nanobots that retire the kidneys, nanoengineered skin, foglets, molecular manufacturing at pennies per pound, and self-replicating probes that eventually colonize the galaxy. Most were timed for the 2020s.

In The Singularity Is Nearer (2024), Kurzweil kept the vision intact: “Nanobots built from diamondoid gears and rotors would be thousands of times faster and stronger than biological materials” (ch. on the Human Body). But in the same chapter, a quieter concession: “There are already devices that can measure blood insulin levels and transfer insulin into the bloodstream, much like a real pancreas.” The Bridge Three vision still centers on diamondoid nanobots. The footnotes now acknowledge the outcome is arriving without them.

Where we actually are

Artificial red cells — the respirocyte, without the diamondoid. Kurzweil wrote in 2005 that “prototypes of respirocytes are still one to two decades in the future” (ch. on the Human Body). The prototypes are here, but they are soft chemistry, not machined carbon. US 12,257,289, granted March 2025 for self-assembling oxygen carriers, describes a lipid-amphiphile vesicle from phospholipids, cholesterol, and encapsulated hemoglobin with an allosteric effector — 2,3-DPG, inositol phosphate, or related regulators — that releases oxygen in a pH-responsive way. Claim 1 reads like a recipe for a pharmacological respirocyte. ErythroMer is the product side: advanced preclinical at the University of Maryland under Allan Doctor, commercialized through KaloCyte, funded at $46M by the DoD. None of it holds your breath for four hours. But it ships oxygen, stores at ambient temperature, and is about to enter humans. Verdict: wrong mechanism, behind on timeline, broadly on trajectory on outcome.

Synthetic platelets — a thousand times faster than what, exactly. Kurzweil predicted “micron-size artificial platelets will achieve homeostasis and bleeding control up to one thousand times faster than biological platelets” by the 2020s (ch. on the Human Body). US patent 10,434,149, granted October 2019 to a Case Western Reserve team, describes a biocompatible flexible nanoparticle, 2-5 μm discoidal in shape with a 10-50 kPa elastic modulus, coated with three classes of peptides: von Willebrand factor-binding, collagen-binding, and active platelet GPIIb-IIIa-binding. The particle homes to injury, grabs collagen and VWF like a biological platelet, and recruits live platelets to amplify clotting. That is the licensed foundation of Haima’s SynthoPlate. In 2025 the company announced DARPA contracts pushing toward an IND in 2027, plus a $1.9M NHLBI Phase II for traumatic brain injury on P2Y12 inhibitors. The “thousand times faster” claim has quietly disappeared — the particle is not faster than a platelet, it is shelf-stable where a platelet is not. A different value proposition than Kurzweil wrote, but the one the market wants. Verdict: wrong mechanism, behind schedule.

Metabolic and digestive nanobots — replaced by pumps, sensors, and peptide drugs. Kurzweil wrote that “by the late 2020s, metabolic nanobots introducing nutrients directly into the bloodstream and guided by wireless body sensors should be reasonably mature” (ch. on the Human Body). They are not, in any diamondoid sense. But the functional outcome arrived via a different stack. The US clinical trial registry lists 157 automated-insulin-delivery studies. Patent grants ramp steadily — eight in 2018, 11 in 2025, eight in the first quarter of 2026 alone. US 12,589,204, granted March 2026, covers artificial-pancreas integrated CGM architectures; US 12,551,619, granted February 2026, describes an AID system using pramlintide alongside insulin. A 2023 NEJM paper on automated insulin delivery in pregnancy complicated by type 1 diabetes (10.1056/nejmoa2303911, 185 citations) showed AID outperforming manual titration in the hardest real-world case. Add GLP-1 agonists that rewrote obesity and diabetes markets from 2022 onward, and you get a closed-loop metabolic stack — a wearable CGM, a pump, a controller algorithm, an injectable peptide — that does most of what the “digestive and metabolic nanobot” was supposed to do. Not through belts or undershirts. Through a subcutaneous cannula and a weekly injection. Verdict: wrong mechanism, outcome largely arriving on Kurzweil’s timeframe.

Elimination nanobots and the retirement of the kidney. Kurzweil wrote that “special elimination nanobots acting like tiny garbage compactors will eventually remove the need for conventional elimination and enable humans to outgrow organs such as the kidneys” by the 2030s (ch. on the Human Body). The Kidney Project at UCSF, led by Shuvo Roy, is building an implantable bioartificial kidney about the size of a coffee cup: a silicon hemofilter that turns blood into ultrafiltrate, and a bioreactor packed with cultured kidney cells that resorbs water and nutrients and concentrates waste into urine. In November 2025 the team reported generating urine in the lab; they are now courting the FDA for feasibility studies and targeting a commercial device by 2030. A 2022 review in Artificial Organs (10.1111/aor.14396) documents the broader wearable/implantable artificial kidney field as real and advancing. None of it is a nanobot. All of it does what the nanobots were supposed to do. Verdict: wrong mechanism, timeline roughly intact for outcome.

Vasculoid and foglet — designs on paper, unchanged. The vasculoid (Freitas’s 500-trillion-nanorobot blood replacement) and foglets (J. Storrs Hall’s nanobot lattice for haptics) existed as engineering sketches in 2005. Twenty-one years later, neither has moved toward a physical prototype. Kurzweil does not mention either in Nearer. Verdict: verified historical; behind on construction.

MNT pennies per pound. Kurzweil predicted molecular manufacturing would drive the cost of any physical product to pennies per pound plus the cost of information (ch. on Work) by the 2020s. Molecular manufacturing does not exist industrially. Additive manufacturing costs have fallen and TSMC/Samsung node transitions keep lowering the cost of a transistor, but neither is atom-by-atom assembly. Verdict: behind schedule.

Self-replicating interstellar probes. Kurzweil wrote that self-replicating nanobot probes would carry civilization beyond the solar system “at speeds very close to the speed of light” (ch. on the Intelligent Destiny of the Cosmos). Breakthrough Starshot, launched 2016 with a pledged $100M, spent roughly $4.5M before (per a 2025 Scientific American account) meetings and funding dried up and the project went quiet. The science survives: in 2025 a Caltech team reported the first direct lab observation of a silicon nitride lightsail pushed by photon momentum. The mechanism is macroscopic foil, not a nanobot swarm. Verdict: behind schedule, wrong mechanism.

Nanoengineered skin. Kurzweil predicted humans would replace their skin with nanoengineered supple materials by the 2030s. The field built electronic skin instead — flexible sensors that sit on the body, not replace it. Highly cited reviews in Advanced Materials (10.1002/adma.202000619, 813 citations; 10.1002/adma.202005902, 577 citations) document the explosion in flexible tactile sensors and conducting hydrogels. E-skin is real. Skin replacement is not. Verdict: wrong mechanism.

“Real reality” body change in the 2040s. Still too early to call.

The scorecard

What Kurzweil missed (and what he nailed)

The pattern is sharp. For most of batch 28, Kurzweil’s outcome prediction is broadly on track — artificial oxygen carriers in preclinical testing, synthetic platelets in IND-enabling studies, implantable bioartificial kidneys heading for the clinic, automated metabolic control in millions of bodies already. But the mechanism is almost uniformly wrong. He predicted diamondoid machines with gears and rotors on a broadcast architecture. What the field built was soft matter — peptide-coated lipid vesicles, self-assembling amphiphiles, silicon hemofilters, GLP-1 analogues, closed-loop algorithms on microcontrollers. Chemistry and control theory won. Diamondoid did not show up.

Forecasting what a technology will do turns out to be easier than forecasting how it will do it. Kurzweil saw the shape of Body 2.0 — a body maintained by programmable, infusable, sensor-guided agents — with real clarity. He just picked the wrong chemistry set. The respirocyte arrived. It is a lipid vesicle.

Method note

Evidence was assembled from a corpus of 9.3M granted US patents with full-text search, a corpus of 357M scientific papers with citation counts and DOIs, the US clinical-trial registry, and web research into named programs including ErythroMer/KaloCyte, Haima Therapeutics’ SynthoPlate, the Kidney Project at UCSF, and Breakthrough Starshot. Every patent number and citation count cited was verified during this session. Both The Singularity Is Near (2005) and The Singularity Is Nearer (2024) were consulted.