Xenobots Explained: A Deep Dive into Programmable Living Machines

Abstract

The emergence of Xenobots—programmable biological machines derived from Xenopus laevis embryos—represents a paradigm shift in the fields of robotics, synthetic biology, and developmental biophysics. First unveiled in 2020 by a multi-institutional team from Tufts University, the University of Vermont (UVM), and Harvard’s Wyss Institute, these constructs challenge the traditional dichotomy between the "born" and the "made." Unlike conventional robots constructed from synthetic materials, Xenobots are composed entirely of living tissue, possessing innate capabilities for self-healing, collective behavior, and, in later iterations, kinematic self-replication. This report provides an exhaustive analysis of the Xenobot program, tracing its evolution from computer-designed heart-muscle actuators (Generation 1.0) to cilia-driven swarms with memory (Generation 2.0) and self-replicating kinematic systems (Generation 3.0). Furthermore, the analysis extends to the recent development of Anthrobots—human-derived biobots with therapeutic potential—and examines the profound ethical, philosophical, and medical implications of engineering living systems from the ground up.

1. Introduction: The Dissolution of the Organism-Machine Boundary

The history of robotics has been dominated by the assembly of non-living materials—metals, plastics, and silicon—into precise, actuated forms controlled by deterministic algorithms. While successful in industrial and computational domains, these "hard" robots lack the adaptability, self-repair mechanisms, and biocompatibility inherent to biological systems.¹ Conversely, developmental biology has traditionally viewed the morphogenesis of organisms as a fixed trajectory dictated by a genomic blueprint. The creation of Xenobots disrupts both narratives, suggesting that the "hardware" of the cell can be decoupled from its evolutionary "software" to create novel, functional life forms that have never existed in nature.³

1.1 Defining the Xenobot

A Xenobot is neither a traditional robot nor a standard animal. It is defined as a "reconfigurable organism" or a "biological machine".⁴ Biologically, it is an aggregate of embryonic cells from the African clawed frog (Xenopus laevis). Functionally, however, it is a robot: a programmable entity designed to perform work. Unlike a genetically modified organism (GMO), a Xenobot contains the standard, unaltered genome of the frog. Its novelty lies not in its DNA, but in its architecture—the specific spatial arrangement of its cells, which dictates its behavior.³

1.2 The Collaborative Triad

The development of Xenobots is the result of a convergence of disciplines across three primary institutions:

1. Tufts University (Allen Discovery Center): Led by Michael Levin, a developmental biologist pioneering the study of bioelectricity and morphological plasticity. Levin’s group focuses on the biological assembly and the manipulation of cellular communication.¹

2. University of Vermont (UVM): Led by Joshua Bongard, a computer scientist and robotics expert. The UVM team utilizes the Deep Green supercomputer to run evolutionary algorithms that design the biological structures in silico before they are built in vivo.¹

3. Harvard Wyss Institute: Contributing expertise in biologically inspired engineering, particularly through the work of Douglas Blackiston (also affiliated with Tufts), who performs the microsurgery and biological assembly.³

This collaboration closes the loop between computational design (the virtual) and biological implementation (the physical), creating a pipeline where artificial intelligence designs the body plan, and biology provides the construction materials.²

2. The Computational Engine: Evolutionary Algorithms and In Silico Design

Before a single cell is harvested, the Xenobot exists as a digital entity. The design process utilizes an "evolutionary algorithm," a form of machine learning inspired by biological natural selection, running on UVM’s Deep Green supercomputer cluster.⁹ This computational phase is critical because the behavior of multicellular aggregates is non-intuitive; a human engineer cannot easily predict how thousands of twitching heart cells or beating cilia will interact to produce directed locomotion.¹¹

2.1 The Evolutionary Pipeline

The algorithm operates through a cycle of simulation, selection, and mutation:

1. Initialization: The computer generates thousands of random 3D configurations of cell types (e.g., passive skin cells and active heart muscle cells).⁵

2. Simulation (The Physics Engine): These designs are placed in a virtual environment that simulates the biophysics of the cells, including friction, fluid dynamics, and intercellular adhesion. The simulation treats the cells like soft-body distinct elements, calculating how the contraction of a muscle cell affects the overall structure.¹⁰

3. Fitness Evaluation: The algorithm assigns a "fitness score" based on a desired task. In early experiments, the task was simple locomotion (distance traveled). In later iterations, it included particle aggregation or phototaxis.¹⁴

4. Selection and Mutation: Designs that perform poorly are discarded ("killed"). High-performing designs are retained and "mutated"—their cell arrangements are slightly altered, or "mated" with other successful designs to produce a new generation.⁵

5. Convergence: After hundreds of generations and months of processing time, the algorithm converges on a set of high-performance designs that are physically realizable.⁹

2.2 Bridging the Sim-to-Real Gap

A major challenge in evolutionary robotics is the "sim-to-real gap"—the discrepancy between the performance of a robot in simulation and the physical world.¹⁷ Biological systems are "noisy" and stochastic; cells do not always behave exactly as modeled. To address this, the UVM team employed "noise injection" and robustness testing. The algorithm was programmed to prefer designs that functioned well even when the simulation parameters (like muscle strength or friction) were randomly fluctuated. This ensured that the final blueprints were robust enough to tolerate the variability of real living tissue.¹⁹

The success of this approach was validated when the in silico predictions matched the in vivo behaviors. When the computer predicted a specific loping gait for a quadrupedal shape, the biological construct exhibited nearly the exact same movement pattern.¹⁹

3. Xenobots 1.0: The Contractile Walkers

The first generation of Xenobots, introduced in early 2020, served as the proof of concept for the entire program. These constructs demonstrated that computer-designed multicellular organisms could be manufactured and retain functionality without a nervous system or external power supply.

3.1 Fabrication Methodology

The construction of Xenobots 1.0 was a labor-intensive, "top-down" process involving microsurgery:

1. Harvesting: Early-stage embryos (blastula stage) of Xenopus laevis were harvested.

2. Dissociation: The embryos were dissociated into single cells.

3. Differentiation: The cells were incubated to differentiate into two specific types: epidermal progenitors (skin) and cardiomyocytes (heart muscle).

4. Assembly: Using the computer-generated blueprints, microsurgeon Douglas Blackiston used tiny forceps and cautery electrodes to manually arrange and sculpt the cells into the specific 3D shapes dictated by the AI. This involved placing heart cells in specific locations to act as motors and skin cells to provide structural passivity.¹

3.2 Mechanism of Action

Movement in Generation 1.0 was driven by the synchronized contraction of heart muscle cells. In an intact frog heart, these cells contract rhythmically to pump blood. In the Xenobot, the AI arranged them such that their contractions were asymmetrical, driving the organism forward or in circles.²¹ The skin cells provided the rigid "chassis" against which the muscles could pull.

3.3 Capabilities and Limitations

These first bots could move coherently, explore their environment, and push small pellets. They relied on embryonic energy stores (yolk platelets) within the cells, allowing them to survive for approximately 7 to 10 days before running out of energy and biodegrading.⁵ However, they had significant limitations: they could not eat, they had limited sensing capabilities, and the manual assembly process was slow and not scalable.¹⁶

4. Xenobots 2.0: Self-Assembly, Cilia, and Memory

The second generation, unveiled in March 2021, marked a significant leap in complexity and manufacturing efficiency. Moving away from the manual "sculpting" of muscle tissue, the team leveraged the intrinsic self-organizing properties of the cells (a "bottom-up" approach) and utilized a different mode of propulsion.¹⁶

4.1 Biological Actuation: The Shift to Cilia

Instead of heart muscles, Xenobots 2.0 utilized cilia—hair-like organelles that protrude from the surface of cells. In adult frogs, cilia line the respiratory tract and are used to push mucus and pathogens out of the lungs. In Xenobots 2.0, the cells were allowed to self-assemble into spheroids. The plasticity of the cells allowed them to repurpose these cilia for locomotion. By beating in coordinated waves, the cilia acted like thousands of tiny oars, propelling the spheroid rapidly through fluid.¹⁶

This shift had three major advantages:

1. Speed: Cilia-driven bots were significantly faster than the muscle-driven precursors.¹⁶

2. Longevity: These bots could survive longer, as the metabolic cost of ciliary beating is generally lower than muscle contraction, and they could navigate diverse environments.¹⁶

3. Scalability: Because they self-assembled into spheroids, thousands could be made in parallel without manual microsurgery.⁶

4.2 Molecular Memory

A groundbreaking feature of Xenobots 2.0 was the introduction of recordable memory. The team injected the stem cells with messenger RNA (mRNA) coding for the photoconvertible protein EosFP. This protein normally glows green. However, upon exposure to a specific wavelength of blue light (390nm), the protein undergoes a conformational change and permanently turns red.¹⁶

In experiments, swarms of Xenobots 2.0 were released into an environment with illuminated zones. Upon retrieval, researchers could count how many bots were glowing red, providing a physical readout of which bots had traveled through the light zone. This demonstrated a "read/write" capability—the environment wrote to the bot's memory, and the scientists read it back. This principle establishes the foundation for future bots that could record exposure to toxins or radioactive materials.¹⁶

4.3 Self-Healing

Both Generation 1.0 and 2.0 displayed remarkable regenerative capabilities. When a Xenobot was sliced nearly in half, the cells detected the injury and migrated to close the gap, restoring the organism's structural integrity and function within 5 minutes. This contrasts sharply with metal robots, where a cut wire or broken chassis results in permanent failure.¹⁶

5. Xenobots 3.0: Kinematic Self-Replication

The most profound development in the Xenobot program, published in late 2021, was the discovery of kinematic self-replication. While all living organisms replicate, they typically do so through growth and splitting (fission) or sexual reproduction (meiosis/gametes). Xenobots 3.0 utilized a mechanism previously observed only in molecules, not in whole organisms.³

5.1 The Mechanism: "Pac-Man" Piling

Kinematic self-replication is a mechanical process. When placed in a dish containing thousands of loose, dissociated stem cells, the parent Xenobots (ciliated spheroids) would swim through the medium. Their movement created currents and collisions that pushed the loose cells together. If enough cells were piled into a cluster (approximately 50 cells), the aggregate would begin to adhere, differentiate, and eventually develop cilia of its own.²³

Once the "offspring" matured (a process taking a few days), they would detach and begin swimming, exhibiting the same behaviors as the parents. They could then push more loose cells into piles, creating grandchildren.³

5.2 AI Optimization and the C-Shape

Initially, spherical Xenobots were inefficient replicators; they could produce one generation, but the lineage would typically die out because the offspring were too small or the piling was inefficient.²⁴ The researchers returned to the Deep Green supercomputer to solve this problem.

The AI simulated billions of body shapes to find one that maximized the efficiency of cell gathering. The result was non-intuitive: a semi-toroid, or "C-shape," bearing a striking resemblance to the video game character Pac-Man.¹¹

• Function: The "mouth" of the Pac-Man shape acted as a scoop, trapping loose cells and compacting them into a ball more effectively than a solid sphere or a flat bulldozer shape could.²⁵

• Result: This AI-designed morphology increased the reproductive output significantly, allowing the system to sustain multiple generations (great-great-grandchildren) before running out of loose cell feedstock.²⁶

5.3 Theoretical Implications

This discovery proved that replication is not solely a function of genomic instructions but can be an emergent property of geometry and mechanics. The cells possessed the potential to replicate in this manner, but that potential was only unlocked when the collective was arranged into the specific C-shape designed by the AI. This validates Levin’s hypothesis regarding the plasticity of the "morphogenetic space"—that the genome is not a strict blueprint but a set of available subroutines that can be rearranged.³

Table 1: Comparative Analysis of Xenobot Generations

Feature	Xenobots 1.0	Xenobots 2.0	Xenobots 3.0
Primary Actuation	Heart Muscle (Contractile)	Cilia (Beating hairs)	Cilia (Beating hairs)
Fabrication	Manual Microsurgery	Self-Assembly (Spheroids)	AI-Optimized Self-Assembly
Morphology	Quadrupedal/Irregular	Spheroidal	C-Shape ("Pac-Man")
Primary Capability	Locomotion, Pushing	Memory, Rapid Swim, Swarming	Kinematic Self-Replication
Design Driver	AI for Locomotion	Biological Self-Organization	AI for Gathering Efficiency
Key Limitation	Slow, non-scalable	Limited manipulation	Requires loose cell feedstock

6. Anthrobots: The Transition to Human Biology

In late 2023, the research team, led by PhD student Gizem Gumuskaya and Michael Levin, demonstrated that the principles derived from Xenopus were not unique to amphibians. They successfully created "Anthrobots"—biological robots derived from adult human cells.⁸

6.1 Biological Source and Construction

Anthrobots are created from human tracheal epithelial cells. In the human airway, these cells possess cilia to move mucus. When extracted and cultured in specific 3D environments (Matrigel), these cells spontaneously form multicellular spheroids (organoids). Unlike Xenobots, which required the AI to design the initial shapes for manual cutting (Gen 1) or specific constraints, Anthrobots self-construct their body plans entirely.⁸

• Inside-Out Turnover: Naturally, these organoids form with cilia on the inside. The researchers developed a chemical protocol to make them turn "inside out," exposing the cilia to the exterior environment, enabling locomotion.³⁰

• No Genetic Modification: Crucially, Anthrobots are wild-type biological structures. They contain no transgenes, making them potentially safer for medical applications.⁸

6.2 The "Superbot" and Neuronal Healing

The most significant finding regarding Anthrobots is their therapeutic utility. In in vitro experiments, researchers grew a 2D layer of human neurons and scratched it to create a "wound." Anthrobots were placed onto the neural tissue.

• Mechanism: The Anthrobots spontaneously aggregated into larger "superbot" chains and settled over the wound.

• Result: Where the Anthrobots were present, the neurons regrew to bridge the gap significantly faster and more robustly than in control groups. The Anthrobots did not just mechanically fill the gap; they induced the neurons to heal themselves, potentially through secreted biochemical factors or mechanotransductive signaling.⁸

This finding is critical because it utilizes patient-specific cells. An Anthrobot made from a patient's own tracheal lining would not trigger an immune response (rejection), paving the way for personalized regenerative bio-machines.⁸

7. Biological Mechanisms: Plasticity and "Junk DNA"

The existence of Xenobots and Anthrobots forces a re-evaluation of fundamental biological concepts, particularly regarding the relationship between the genome and the phenotype.

7.1 Teleophobia and Goal-Directedness

Michael Levin argues that biology is "teleophobic"—scientists are often uncomfortable attributing goal-directedness to cells. However, Xenobot research suggests that cells possess an innate "collective intelligence" or plasticity.³ When liberated from the restrictive context of the embryo (where they are "told" to build a frog), they do not die or become unstructured tumors. Instead, they cooperate to build a new life form with a different homeostasis and different behaviors (e.g., kinematic replication).

7.2 The Role of "Junk DNA"

Critics often assume that non-coding regions of DNA ("junk DNA") have no function. However, the Xenobot platform suggests that this DNA may encode for alternative body plans or "subroutines" that are never activated in the standard environment but can be accessed under novel conditions. The genome is not a blueprint for a specific animal, but a toolkit for building various multicellular machines depending on the boundary conditions.³²

7.3 Stem Cell Aggregation Dynamics

The ability of these cells to aggregate is driven by surface proteins and physical forces. The "stickiness" of the cells, combined with the AI-optimized geometries (like the Pac-Man mouth), exploits the physics of soft matter. The replication observed in Xenobots 3.0 is a mechanical phenomenon facilitated by biological adhesion—a hybrid of physics and biology.²¹

8. Applications: The Future of Bio-Robotics

The transition from theoretical biology to applied technology is the ultimate goal of the Xenobot/Anthrobot program. The potential applications are vast, owing to the biodegradability and biocompatibility of the systems.

8.1 Regenerative Medicine

The primary application domain is medicine. Because Anthrobots can be made from a patient's own cells:

• Arterial Scrubbing: Biobots could be deployed into the circulatory system to scrape plaque from arterial walls, reducing the risk of atherosclerosis.¹²

• Targeted Drug Delivery: Biobots could carry pharmaceutical payloads to specific tumor sites. Unlike nanoparticle vectors, which are passive, biobots could actively swim against blood flow or navigate complex tissue architectures.³⁴

• Tissue Repair: As evidenced by the neural healing experiments, biobots could act as "living sutures" or pro-regenerative patches for spinal cord injuries or retinal damage.⁸

8.2 Environmental Remediation

Xenobots are biodegradable; when they die, they simply decompose into protein. This makes them ideal for environmental release, unlike plastic or metal robots which contribute to pollution.

• Microplastic Collection: Swarms of Xenobots could be deployed in waterways to aggregate microplastics into large balls that can be easily harvested.¹²

• Radioactive Cleanup: Biobots could be engineered (or evolved) to sense and gather radioactive contaminants in areas too small or dangerous for humans or standard robots.⁹

8.3 Biological Sensing

Leveraging the molecular memory capabilities demonstrated in Xenobots 2.0, these organisms could serve as sentinel swarms. Released into a water supply or a patient's body, they could record the presence of specific pathogens, toxins, or chemical signatures, providing a detailed map of contamination or disease upon retrieval.¹⁶

9. Ethical, Legal, and Social Implications (ELSI)

The creation of self-replicating, programmable living systems raises significant ethical questions. While the researchers emphasize that these entities have no brains, nervous systems, or capacity for suffering, the "optics" of living robots create public concern.³⁶

9.1 The Definition of Life

Xenobots challenge legal and ethical definitions. Are they animals? If so, do they require review by IACUC (Institutional Animal Care and Use Committee)? Currently, because they are embryonic and lack a nervous system, they fall outside many animal welfare regulations, but their status as "living" entities creates a grey area.⁴

9.2 The "Grey Goo" Scenario

The announcement of self-replicating robots triggered fears of a "grey goo" scenario—robots replicating uncontrollably and consuming all resources. However, the researchers note that Xenobot replication is strictly "kinematic." It requires a specific feedstock (loose stem cells) that does not exist in nature outside of a lab dish. Without a supply of dissociated frog cells, Xenobots cannot replicate. They are thermodynamically open systems that cannot sustain themselves in the wild.³

9.3 Dual-Use and Weaponization

Like any technology, there is a risk of dual use. Biological machines designed to carry drugs could theoretically be designed to carry toxins or bioweapons. The "democratization" of this technology—where an AI and a standard biology lab are sufficient to create novel life forms—necessitates strict governance and ethical oversight.³⁹

9.4 "Playing God" and the Yuck Factor

Public response often cites the "unnatural" nature of these creations. Ethicists argue that this reaction (the "yuck factor") must be balanced against the potential benefits. If Anthrobots can cure spinal cord paralysis or clear cancer without side effects, the moral imperative may tilt toward their development, provided safety protocols are robust.³³

10. Conclusion

Xenobots and Anthrobots represent more than just a novelty or a niche sub-field of robotics; they are the harbingers of a new era in which biology is treated as a programmable medium. By closing the loop between evolutionary AI and developmental biology, researchers at Tufts, UVM, and Harvard have demonstrated that the "body plan" is not an immutable destiny written in DNA, but a flexible construct that can be manipulated to serve human needs.

The progression from the contractile, hand-built Xenobots 1.0 to the self-replicating, AI-optimized Xenobots 3.0, and finally to the medically relevant human Anthrobots, illustrates a rapid maturation of the technology. As we move into 2025 and beyond, the focus will shift from "can we build it?" to "how can we control it?" and "what should we cure with it?".³⁹

The implications extend far beyond the laboratory. They challenge our understanding of evolution, offering a view of life where intelligence and form are plastic, reconfigurable, and, ultimately, designable. As Michael Levin posits, we are witnessing the birth of a new discipline: the engineering of synthetic morphology, where the distinction between the machine and the organism finally disappears.

Works cited

1. Living Machines Are Created in the Lab | Tufts Now, accessed January 2, 2026, https://now.tufts.edu/2020/01/16/living-machines-are-created-lab

2. AI Xenobots - Communications of the ACM, accessed January 2, 2026, https://cacm.acm.org/news/ai-xenobots/

3. Team builds first living robots—that can reproduce - Wyss Institute, accessed January 2, 2026, https://wyss.harvard.edu/news/team-builds-first-living-robots-that-can-reproduce/

4. Biological Robots: Perspectives on an Emerging Interdisciplinary Field - PMC, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC10442684/

5. Not bot, not beast: scientists create first ever living, programmable organism - Find an Expert, accessed January 2, 2026, https://findanexpert.unimelb.edu.au/news/9267-not-bot--not-beast--scientists-create-first-ever-living--programmable-organism

6. Xenobots 2.0: These Living Robots Self-Assemble From Frog Skin Cells - SingularityHub, accessed January 2, 2026, https://singularityhub.com/2021/04/02/scientists-made-living-self-healing-robots-from-frog-skin-cells/

7. XENOBOTS- LIVING ROBOTS THAT CAN REPRODUCE - PROF JOSH BONGARD, accessed January 2, 2026, https://www.youtube.com/watch?v=9JmyrGeY5z0

8. Scientists build tiny biological robots from human cells - Wyss Institute, accessed January 2, 2026, https://wyss.harvard.edu/news/scientists-build-tiny-biological-robots-from-human-cells/

9. Team Builds the First Living Robots - University of Vermont, accessed January 2, 2026, https://www.uvm.edu/uvmnews/news/team-builds-first-living-robots

10. XENOBOT: – ExpressedToUnfold - WordPress.com, accessed January 2, 2026, https://expressedtounfold.wordpress.com/2020/06/19/xenobot/

11. AI-Designed Self-Replicating "Xenobots," Shaped Like Pac-Man, Are Science Fiction No Longer - Hackster.io, accessed January 2, 2026, https://www.hackster.io/news/ai-designed-self-replicating-xenobots-shaped-like-pac-man-are-science-fiction-no-longer-56b23f438cf4

12. Artificial lifeforms designed by supercomputers are fully programmable - New Atlas, accessed January 2, 2026, https://newatlas.com/biology/programmable-artificial-lifeforms-xenobots/

13. A scalable pipeline for designing reconfigurable organisms | PNAS, accessed January 2, 2026, https://www.pnas.org/doi/10.1073/pnas.1910837117

14. Investigating an A-star Algorithm-based Fitness Function for Mobile Robot Evolution - Greenwich Academic Literature Archive (GALA), accessed January 2, 2026, https://gala.gre.ac.uk/id/eprint/21052/7/21052%20PRABHU_Investigating_an_A-star_Algorithm-Based_Fitness_Function_2018.pdf

15. Xenobots: Bioengineered Living Machines For The Future Using Artificial Intelligence - IJCRT.org, accessed January 2, 2026, https://ijcrt.org/papers/IJCRT2306768.pdf

16. Scientists Create the Next Generation of Living Robots - Tufts Now, accessed January 2, 2026, https://now.tufts.edu/2021/03/31/scientists-create-next-generation-living-robots

17. Bridging the Sim-to-Real Gap with Dynamic Compliance Tuning for Industrial Insertion, accessed January 2, 2026, https://www.research.autodesk.com/publications/sim-to-real-gap-dynamic-compliance-tuning-industrial-insertion/

18. Bridging the Sim-to-Real Gap in Humanoid Dynamics via Learned Nonlinear Operators - OpenReview, accessed January 2, 2026, https://openreview.net/pdf?id=Zi1oawm4cA

19. The Emergence of Xenobots: From Code to Creature | The Scientist, accessed January 2, 2026, https://www.the-scientist.com/from-code-to-creature-71877

20. Crossing the Gap: A Deep Dive into Zero-Shot Sim-to-Real Transfer for Dynamics, accessed January 2, 2026, https://www.robot-learning.uk/crossing-the-gap

21. Xenobots: Living Robots that Make Science Fiction a Reality - Lab Manager, accessed January 2, 2026, https://www.labmanager.com/xenobots-living-robots-that-make-science-fiction-a-reality-30661

22. The next generation of living machines: xenobots 2.0 - Freethink Media, accessed January 2, 2026, https://www.freethink.com/science/xenobots

23. Tiny living machines called xenobots can create copies of themselves - Science News, accessed January 2, 2026, https://www.sciencenews.org/article/tiny-living-machines-xenobots-replicate-copies-frog-cells

24. Scientists Unveiled the World's First Living Robots Last Year. Now, They Can Reproduce, accessed January 2, 2026, https://www.smithsonianmag.com/smart-news/scientists-unveiled-the-worlds-first-living-robots-last-year-now-they-can-now-reproduce-180979150/

25. Living robots made in a lab have found a new way to self-replicate, researchers say | WUNC, accessed January 2, 2026, https://www.wunc.org/2021-12-01/living-robots-made-in-a-lab-have-found-a-new-way-to-self-replicate-researchers-say

26. Kinematic self-replication in reconfigurable organisms - PNAS, accessed January 2, 2026, https://www.pnas.org/doi/10.1073/pnas.2112672118

27. Cells Form Into 'Xenobots' on Their Own - Quanta Magazine, accessed January 2, 2026, https://www.quantamagazine.org/cells-form-into-xenobots-on-their-own-20210331/

28. Scientists Build Tiny Biological Robots from Human Cells - Tufts Now, accessed January 2, 2026, https://now.tufts.edu/2023/11/30/scientists-build-tiny-biological-robots-human-cells

29. From Xenobots to Anthrobots - Kim Bellard - Medium, accessed January 2, 2026, https://kimbellard.medium.com/from-xenobots-to-anthrobots-5b87b55c703e

30. Anthrobots: age reversal, ancient genes, and what new beings are telling us about genetics and evolution, accessed January 2, 2026, https://thoughtforms.life/anthrobots-age-reversal-ancient-genes-and-what-new-beings-are-telling-us-about-genetics-and-evolution/

31. Developing biological robots to repair damaged neural tissue - RegMedNet, accessed January 2, 2026, https://www.regmednet.com/developing-biological-robots-to-repair-damaged-neural-tissue/

32. [BONUS] Xenobots and rethinking 'junk DNA': Tiny Show and Tell Us #12, accessed January 2, 2026, https://www.acs.org/pressroom/tiny-matters/xenobots-and-junk-dna.html

33. XenoBots (living robots) – Medical revolution or Ethical mindfield? - The House of Ethics, accessed January 2, 2026, https://www.houseofethics.lu/2021/08/12/xenobots-medical-revolution-or-ethical-mindfield/

34. The Living Robots: Xenobots & Biotechnology - Chemical Express Thailand, accessed January 2, 2026, https://chemicalexpressth.com/articles/xenobots/

35. What are Xenobots? - ClinicalPro, accessed January 2, 2026, https://www.clinicalpro.co.uk/what-are-xenobots/

36. The Potential of Xenobots, a Living Machine - Hadron, accessed January 2, 2026, https://sites.imsa.edu/hadron/2024/04/02/the-potential-of-xenobots-a-living-machine/

37. “Living Robots”: Ethical Questions About Xenobots | Request PDF - ResearchGate, accessed January 2, 2026, https://www.researchgate.net/publication/341139092_Living_Robots_Ethical_Questions_About_Xenobots

38. Ethics and responsibility in biohybrid robotics research - PMC, accessed January 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC11294997/

39. Xenobots - AI for Good - ITU, accessed January 2, 2026, https://aiforgood.itu.int/speaker/xenobots/

40. Blog - “Living robots”: Ethical questions about Xenobots - Bioethics Today, accessed January 2, 2026, https://bioethicstoday.org/blog/living-robots-ethical-questions-about-xenobots/

41. With Living Robots, Scientists Unlock Cells' Power to Heal | Tufts Now, accessed January 2, 2026, https://now.tufts.edu/2024/03/22/living-robots-scientists-unlock-cells-power-heal