Xenobot
Frog skin cells — given no instructions — assembled themselves into tiny swimmers that can replicate.
Xenobots are sub-millimeter (~500–1,000 µm) self-organized aggregates of dissociated Xenopus laevis embryonic cells — typically pluripotent skin-progenitor cells, sometimes combined with cardiac cells whose contractions provide propulsion or with cilia-bearing skin cells whose beating cilia drive swimming. Designed in silico by an evolutionary algorithm running a soft-body physics simulator, then sculpted from cell aggregates by microsurgery (Levin & Bongard 2020), xenobots exhibit collective locomotion, wound healing, and — in 2021 — kinematic self-replication: a swimming xenobot pushes loose stem cells into piles that mature into new xenobots over ~5 days. Anthrobots (Levin lab 2023) use adult human tracheal epithelial cells with similar self-organizing behaviors.
The 2020 Levin–Bongard pipeline begins with an evolutionary search in a soft-body physics simulator: thousands of in-silico body plans built from contractile and passive 'cells' compete on a fitness function (e.g., locomotion speed). Top designs are translated to physical xenobots by harvesting Xenopus stage-9 embryonic ectoderm and cardiac mesoderm; cells are dissociated, allowed to re-aggregate, then microsurgically sculpted with electrocautery into the simulator's recommended morphology. The aggregates run on the chemical energy stored in their yolk and survive ~7–14 days in pond water at room temperature. The 2021 self-replication paper showed that spheroidal xenobots, tagged with red fluorophore, swam in circles inside a dish containing ~50,000 loose stem cells; over hours, they pushed cells into piles that, given another ~5 days, matured into new motile spheroids — a kinematic, not genetic, replication mode. Anthrobots (2023) use adult human airway-epithelial cells grown into ciliated organoids that self-organize without genetic modification, and have been observed bridging gaps between dishes of cultured neurons.
In plain terms
Imagine you took a few thousand cells from a frog egg and gently let them clump together in a dish. By themselves, with no brain and no instructions, they decide to grow tiny hairs and start swimming around. If you cut one in half, both halves keep going. And if you put loose cells in their pond, the swimmers nudge them into piles that, after a few days, become more swimmers. They are tiny, alive, and surprisingly purposeful — but they are not animals; they are biological robots made from animal parts.
Also called: frog-cell biorobot, anthrobot (human-cell variant).
Why it matters
- Demonstrates that morphological autonomy — body-plan formation, locomotion, replication — emerges from generic biological substrate plus physical constraints, not from species-specific genetics.
- Reframes 'robot': a biological agent that is alive, self-healing, and biodegradable, with no electronics or synthetic materials.
- Provides a model system to test theories of biological agency and basal cognition advanced by Michael Levin.
- Anthrobots from a patient's own cells could one day bridge severed neural tissue or clear arterial plaque without immune rejection — a long-horizon medical hypothesis.
Where it lives
- Levin Lab at Tufts (xenobots, anthrobots)
- Bongard Lab at U. Vermont (evolutionary design)
- Sam Kriegman simulator pipeline
Key numbers
- Size: ~500–1,000 µm (Sub-millimeter, visible under a stereoscope)
- Lifespan: ~7–14 days (Yolk-fueled; no external feeding required)
- Kinematic replication cycle: ~5 days (Levin & Bongard, PNAS 2021)
- Cell source: Xenopus stage-9 embryos (Anthrobots use adult human tracheal cells)
Surprising facts
- Xenobots heal wounds: cut one in half and both pieces re-shape into smaller, still-motile xenobots within hours, with no genetic engineering — the cells simply re-run their morphogenetic program.
- Spheroidal xenobots demonstrated kinematic self-replication in 2021: by pushing loose stem cells in a dish, a parent xenobot induces formation of daughter xenobots — a non-genetic mode of reproduction not previously observed in animal cells.
- Anthrobots — built from a single donor's adult human tracheal cells, no genetic modification — formed motile multicellular structures and bridged gaps in cultured neurons, in laboratory dishes (Levin lab 2023).
Open questions
- Self-replication semantics are contested: The 2021 'kinematic self-replication' is a piling-up behavior in which parent xenobots push loose cells into clumps that — given the same favorable environment — become daughter xenobots. Critics note that this requires an exogenous reservoir of cells and is unlike genetic replication; the authors agree, terming it kinematic rather than genetic. Whether to call it 'replication' is a definitional debate, not a measurement dispute.
- Anthrobot results are single-lab and unreplicated: The 2023 anthrobot paper reports striking behaviors — bridging cuts in cultured neurons — but is currently a single-laboratory result; independent replication has not yet been published as of mid-2026. The biological mechanism for cross-dish bridging is not established.
A short history
Michael Levin (Tufts) and Josh Bongard (U. Vermont) introduced the term 'xenobot' in PNAS 2020, presenting an evolutionary-design pipeline that produces motile millimeter-scale aggregates of Xenopus cells. A January 2021 follow-up reported kinematic self-replication. In late 2023 the Levin lab announced 'anthrobots' — analogous self-assembling structures from human tracheal epithelium — extending the framework to adult mammalian cells without genetic modification. The work is part of Levin's broader research program on basal cognition and morphogenetic agency.
Sources
- Kriegman, Blackiston, Levin, Bongard — PNAS (2020) (https://www.pnas.org/doi/10.1073/pnas.1910837117)
- Kriegman et al., PNAS (2021) — kinematic self-replication (https://www.pnas.org/doi/10.1073/pnas.2112672118)
- Gumuskaya et al., Advanced Science (2023) — anthrobots (https://onlinelibrary.wiley.com/doi/10.1002/advs.202303575)
- Levin Lab homepage (https://drmichaellevin.org/)