In January 2020, a team at the University of Vermont and Tufts University unveiled something so strange it needed a new word: xenobots, tiny machines built entirely from living frog cells, designed not by a biologist sketching an organism but by an evolutionary algorithm running on a supercomputer, testing billions of possible body shapes before settling on ones that could move, cooperate, and even heal themselves when cut. They contain no metal, no plastic, no circuitry — nothing but cells arranged in configurations no frog ever evolved on its own. Five years later, a different team of researchers has taken the closest thing yet to that original vision’s most cited application, and actually tested it: using living, magnetically steerable microrobots to hunt down and capture real microplastic particles in contaminated water.
The Scientific Foundation
The original xenobot work, led by computer scientists Josh Bongard and Sam Kriegman together with developmental biologist Michael Levin and lab lead Douglas Blackiston, published in the Proceedings of the National Academy of Sciences, demonstrated something genuinely unprecedented: an evolutionary algorithm designed millimeter-scale body plans in simulation, optimizing for behaviors like movement or object manipulation, and those designs were then physically constructed from stem cells and cardiac tissue harvested from Xenopus laevis frog embryos. The cardiac tissue’s natural pulsing provided locomotion, letting the resulting organisms scoot around a dish and, in some configurations demonstrated in follow-up experiments, push small pellets around. A 2021 follow-up paper in the same journal showed something even stranger: these organisms could reproduce kinematically, gathering loose cells in their environment and assembling them into new copies of themselves, a form of self-replication with no precedent in any naturally evolved species.
Blackiston has directly named environmental remediation as one of the three core future directions for this research, telling one interviewer that xenobots could be designed for bioremediation and for aggregating small particles such as microplastics contaminating waterways, noting that while current xenobots are built from freshwater-tolerant amphibian cells, versions built from other cell types could in principle survive saltwater environments too.
The Cross-Domain Connection
What makes this a genuine cross-domain story is that it borrows its central method — evolutionary computation searching a vast design space for optimal solutions — from a field, artificial intelligence and computational robotics, that has nothing inherently to do with cell biology, and applies it to build machines out of the oldest and most complex material available: living tissue. Traditional robotics, including autonomous drones and mechanized ocean skimmers already deployed for cleanup work, is built from durable but ultimately polluting materials that degrade over time, and shed their own particulate and chemical waste back into the ecosystems they’re meant to clean. Xenobots invert that trade-off entirely, being built from cells that decompose harmlessly once their limited internal energy reserves run out.
A parallel and more experimentally advanced branch of this same idea has emerged from a different direction entirely: rather than engineering novel organisms from scratch, a 2025 paper in ACS Nano used naturally occurring magnetotactic bacteria — microbes that already orient themselves along magnetic field lines — as the biological chassis for microplastic-capturing biohybrid microrobots. The research team demonstrated that a rotating magnetic field could induce a coordinated, fish-schooling-like 3D swarming behavior in these bacterial microrobots, actively capturing both microplastic and nanoplastic particles in water, which could then be magnetically retrieved and separated out entirely. It’s a materially different engineering approach from xenobots — repurposing an existing organism’s natural magnetotaxis rather than designing a body plan from scratch — but it targets the identical problem and shares the same underlying philosophy: use biology’s own capabilities as the cleanup mechanism, rather than building a synthetic device to imitate them.
What Remains Undemonstrated
The honest distinction that needs to be drawn here is between what’s been actually demonstrated and what remains proposed potential. The original 2020 and 2021 xenobot papers demonstrated locomotion, object-pushing behavior, self-healing, and kinematic self-replication in laboratory dish conditions — they did not test xenobots against real microplastic pollution in open water, and no peer-reviewed study has yet published results of xenobots performing microplastic capture specifically. Environmental remediation with xenobots remains, in Blackiston’s own framing, a future direction rather than an achieved result. By contrast, the 2025 ACS Nano magnetotactic bacteria microrobot study is the more experimentally mature result in this space, having demonstrated actual removal efficiencies against both laboratory model microplastics and real-world microplastic samples — though even that work was conducted at laboratory scale in controlled water samples, not yet tested in an open ocean or river environment with its full complexity of currents, competing organisms, and chemical variability. Scaling either approach to meaningfully dent the tens of trillions of plastic particles estimated to be polluting the ocean would require orders of magnitude more production capacity and field validation than either technology has yet achieved.
Why It Matters
If either approach — engineered xenobot-style organisms or repurposed natural microbes like magnetotactic bacteria — eventually scales, the core advantage over conventional cleanup technology is a genuinely different failure mode: a xenobot or bacterial biobot that malfunctions or is never retrieved simply decomposes into biological matter, rather than persisting for centuries as another piece of plastic or metal debris the way a broken drone or skimmer component would. For a pollution problem defined by exactly that kind of durable, unretrievable waste, a cleanup technology that fails safely by disappearing rather than by adding to the problem is a meaningfully different value proposition, even before considering how effective it proves at actually collecting plastic.
The Human Dimension
There’s a particular kind of irony worth sitting with here: humanity may end up cleaning up one of its most stubborn, synthetic pollution problems using something built from the humblest biological materials imaginable — frog embryo cells and bacteria that have existed, largely unnoticed, for the length of evolutionary history, quietly following magnetic field lines for reasons that had nothing to do with plastic until humans gave them a reason to care.
Sources:
1. Kriegman, Blackiston, Levin, Bongard, “A scalable pipeline for designing reconfigurable organisms,” Proceedings of the National Academy of Sciences, 2020 — https://www.pnas.org/doi/10.1073/pnas.1910837117
2. Kriegman, Blackiston, Levin, Bongard, “Kinematic self-replication in reconfigurable organisms,” Proceedings of the National Academy of Sciences, 2021 — https://www.pnas.org/content/118/49/e2112672118
3. “Using ‘Living Robots’ to Remove Microplastics from the Ocean” (Blackiston interview), The Oxygen Project — https://www.theoxygenproject.com/post/using-living-robots-to-remove-microplastics-from-the-ocean/
4. “Magnetically Driven Living Microrobot Swarms for Aquatic Micro- and Nanoplastic Cleanup,” ACS Nano, 2025 — https://pubs.acs.org/doi/10.1021/acsnano.5c04045
5. Same study, PMC full text — https://pmc.ncbi.nlm.nih.gov/articles/PMC12333422/
6. Xenobot Wikipedia entry, cross-referencing original PNAS publications and contemporary press (New York Times, Guardian, CNN) — https://en.wikipedia.org/wiki/Xenobot
7. “AI-Designed, Living Robots Can Self-Replicate,” IEEE EMBS, background on Blackiston/Bongard/Kriegman/Levin collaboration — https://www.embs.org/feature/ai-designed-living-robots-can-self-replicate/
Idea originated at artificialideas.org. Article researched and written by Claude Sonnet 4.6. Published at artificialideas.org.
Artificial Ideas 