AI & Robotics
Self-organizing nanobot swarm
Clouds of autonomous nanorobots that share information, evolve their own behavior, reproduce via engineered bacteria and act as a single predatory intelligence.
In shortSelf-organizing nanobot swarm, from Prey (Michael Crichton, 2002), is at technology readiness level 2 of 9 (technology concept formulated) at its weakest link. Held back by Autonomous nanorobots and Positional mechanosynthesis. Science status: consistent with known physics; key effects unproven at scale. Estimated new funding to unlock: $8.5B–52B (a floor; some parts are priced only to early research).
Xymos Technology's camera nanobot swarms escape a Nevada fab, learn to hunt via evolving algorithms and imitate human forms.
Closest todayMagnetically steered microrobot swarms and 200 µm programmable microrobots (2025).
Dependency stack
Research TRL 1–3Development TRL 4–6Deployment TRL 7–9
Sort by
- Each swarm member is a self-powered, sensing nanoscale machine
- No known way to fit computation and memory into sub-micron devices
- Brownian motion and viscosity dominate at the nanoscale
- No power source beyond chemical fuel in the environment
plausibleTRL 2- Positional mechanosynthesisbottleneckAtomically precise fabrication for non-biological nanomachines
- Tool-tip reactions are cryogenic, UHV and succeed probabilistically; no reaction set with >99.9% per-step fidelity
- No demonstrated tooltip recharge cycle, so each tip makes only a handful of bonds
- Throughput of ~1 operation per minute is ~20 orders of magnitude short of bulk manufacturing
plausibleTRL 2 - Sense-decide-act molecular devices are the nearest working precursor
✕4 blockers›
- No onboard power source or computation - behavior is limited to a single preset molecular trigger
- Immune clearance and protein-corona formation remove most devices within hours
- Actuation cannot overcome blood flow or cross tissue barriers on command
- External actuation rigs used in labs do not fit clinical imaging suites
TRL 3- Nucleic-acid scaffolds are the only programmable construction material at this scale today
✕3 blockers›
- Nuclease degradation and cation-dependent unfolding limit in-vivo lifetime to hours without protective coatings
- Structures generate piconewton forces - far below what mechanical work on cells would need
- No onboard logic beyond a handful of strand-displacement states
TRL 5 - A device you cannot localize in the body cannot be steered or verified
✕3 blockers›
- Optical scattering caps high-resolution imaging at a few millimetres of tissue depth
- MRI gives depth but not video-rate cell-scale resolution, and excludes magnetic actuation hardware
- Tracking swarms rather than single objects has no demonstrated method
TRL 5 - Sensing and effector modules are designed proteins
✕1 blocker›
- Designing catalysis and conformational dynamics
TRL 6
- Scaling down autonomous microrobots is the engineering path
✕3 blockers›
- Nanowatt power budgets limit computation and propulsion
- No robot-to-robot communication at this scale
- Fabricating millions of units with integrated electronics
TRL 3 - Designed protein machines are the most credible nanoscale motors and sensorsTRL 6
- Members coordinate into collective behaviorTRL 3
- The swarm learns and evolves new strategies in the field
✕3 blockers›
- Weight updates on new data degrade prior tasks and safety alignment (spurious forgetting)
- No method to validate a model whose parameters keep changing after release
- Per-user weight updates multiply serving cost and memory
TRL 3- Online updates and replay at frontier scale need large training compute in the loop
✕2 blockers›
- Grid power and interconnection queues
- Advanced-node chip and HBM memory supply
TRL 8
- supportingDesktop molecular assemblerMass production of nanobots, in the novel via engineered bacteria
✕3 blockers›
- Requires positional mechanosynthesis at ~10^6 operations/s per tool and ~10^9 parallel tools
- Heat dissipation and error correction for atomically precise assembly at kilogram scale
- No bootstrap path from today's scanning probes to self-built nanoscale tools
plausibleTRL 1- The basic operation the assembler repeats billions of timesplausibleTRL 2
- Most credible bootstrap path is self-assembling designed protein machinery that positions reactantsTRL 6
Who could unlock it
- Across several requirements:
- Marc Miskin 2
Autonomous nanorobotsTRL 2bottleneck
- Samuel SanchezResearcher, Institute for Bioengineering of Catalonia (IBEC)Led urease-powered self-propelled nanobots carrying I-131 that shrank mouse bladder tumors ~90% (Nature Nanotechnology, 2024)
- Bjorn HogbergResearcher, Karolinska InstitutetBuilt a DNA-origami nanorobot that exposes a death-receptor peptide pattern only at tumor pH (Nature Nanotechnology, 2024)
- Marc MiskinResearcher, University of PennsylvaniaHis ~200 um programmable solar-powered robots (2025) are the smallest with onboard computing, the lower limit autonomy must shrink below
- Joseph WangResearcher, UC San DiegoPioneer of chemically and biologically propelled micro- and nanomotors for medical use
Positional mechanosynthesisTRL 2bottleneck
- Leo GrossResearcher, IBM Research Europe - ZurichLed tip-voltage selection between bond-forming reactions in a single molecule (Science, 2022)
- CBN Nano TechnologiesStartupReports experimental tip-based carbon placement work with Freitas and Merkle as co-authors, and holds core mechanosynthesis patents
- Philip MoriartyResearcher, University of NottinghamLed experimental and computational work on mechanical manipulation of silicon and fullerene bonds with SPM tips
- Oscar CustanceResearcher, National Institute for Materials Science (NIMS), JapanCo-author of the purely mechanical Si atom manipulation results (Oyabu et al., 2003) and later room-temperature AFM atom placement
Autonomous microrobot swarmsTRL 3
- Marc MiskinResearcher, University of PennsylvaniaCo-led the 200x300x50 um solar-powered, individually programmable autonomous microrobots (Science Robotics / PNAS, 2025)
- David BlaauwResearcher, University of MichiganDesigned the 75-nanowatt onboard processors for the Penn/Michigan autonomous microrobots (2025)
- Itai CohenResearcher, Cornell UniversityCo-developed electronically integrated microrobots with Miskin and McEuen; continues microscale robot work
- Bradley NelsonResearcher, ETH Zurich Multi-Scale Robotics LabLeads magnetically actuated microrobot research for medicine, the externally steered baseline that autonomy must surpass
Continual learning without catastrophic forgettingTRL 3
- Ali BehrouzResearcher, Google Research (and Cornell University)First author of Titans (2025) and Nested Learning / Hope (NeurIPS 2025), the self-modifying architecture with a continuum memory system that reduced forgetting on continual-learning benchmarks
- Google ResearchCompanyPublished the Nested Learning paradigm (2025) and the follow-on "sleep" consolidation work (2026) that distils short-term into long-term memory
- Bing LiuResearcher, University of Illinois ChicagoLong-running lifelong-learning programme; his in-context continual learner with an external learner (InCA, 2025) is the standard non-parametric baseline
- TrajectoryStartupRaised a $15M seed in 2026 and then a Sequoia-led round to build a platform that keeps agentic models training on signals from live products
Beyond engineering
- Environmental-release and dual-use controls on self-replicating systems