Weekly Piece of Future #161
From Artificial Neurons to Cancer-Beating Dogs and Robots That Read Your Mind
Hey there, fellow future-addicts!
Welcome to this week's edition of Rushing Robotics! Another week, another batch of breakthroughs that make you do a double-take. This week we've got artificial neurons speaking the brain's language, a dog beating cancer with a custom mRNA vaccine, robots reading your mind in real time, and an AI running from space to control a humanoid on Earth. If that sounds like a lot — it is. Each of these stories alone would have been the headline of the decade a generation ago. This week, they're all in the same newsletter. And we're just getting started.
🤯 Mind-Blowing
Researchers at UMass Amherst built the first artificial neuron capable of communicating directly with living biological tissue, operating at voltages that actually match the human brain. A rescue dog in Sydney received the world's first personalized mRNA cancer vaccine, designed with help from ChatGPT and a chain of researchers connected through AI-assisted genomic analysis. Oklahoma State University demonstrated robots that respond to your brain's error signals before your hands can even react. China approved the world's first commercially authorized brain-computer interface. And a humanoid robot was controlled by an AI model running entirely from satellites in low Earth orbit.
🔊 Industry Insights & Updates
The robotics and deep-tech industries are entering a new phase — defined less by proof-of-concept demos and more by the hard work of actually scaling. A South Korean startup launched a 20-DoF robotic hand weighing under a kilogram, built to integrate into real humanoid platforms. Techman Robot unveiled a factory-floor humanoid at NVIDIA GTC 2026, targeting commercialization before year's end. UBTech and Siemens committed to producing 10,000 humanoid robots annually — not a roadmap target, but a 2026 commitment. And Humboldt University Berlin cracked a key bottleneck in building a practical quantum internet using diamonds.
🧬 BioTech
Biology isn't just being studied anymore — it's being engineered and put to work. The University of Edinburgh figured out how to turn discarded PET plastic bottles into L-DOPA, the primary Parkinson's medication, using genetically modified bacteria. Rice University unveiled a living bacteria-powered sensor that detects chemicals through the electrical signals of microbes — with applications across food safety, wastewater, and industrial monitoring. And a randomized controlled trial confirmed that stroke patients using an at-home, non-invasive BCI recovered meaningful hand function at rates that shattered long-held assumptions — 55.5% responded meaningfully versus just 9.6% in the control group.
💡 Products/Tools of the Week
🎥 Video Section
This week's picks capture the energy of the moment perfectly — Jensen Huang sharing the stage with a real humanoid at NVIDIA GTC, China's fully autonomous tennis-playing robot, and the Phantom MK-1 combat drone already being field-tested in Ukraine. Three different windows into where physical AI is heading, and how fast it's getting there.
Every week I sit down to write this and think - surely it can't keep accelerating at this pace. And every week, the stories prove me wrong. The convergence of AI, biotech, neuroscience, and robotics is no longer a prediction; it's the present tense. What excites me most isn't what landed this week, it's knowing that somewhere right now, the next breakthrough is already in a lab, waiting to show up in your inbox. Stay hungry, stay futurish!
🤯 Mind-Blowing
The first artificial neuron capable of communicating directly with the human brain has been built by Jun Yao and colleagues at the University of Massachusetts Amherst. The device operates at approximately 0.1 volts — matching the 70 to 130 millivolt range of real neurons — while previous artificial neurons required up to 0.5 volts and consumed 100 times more power, making biological interaction impossible. A memristor tuned with protein nanowires from Geobacter sulfurreducens bacteria sits at the core, switching on near 60 millivolts and self-resetting after each firing to mimic the natural pulse-and-pause rhythm of neural activity. UMass Amherst researchers also demonstrated chemical responsiveness, with sodium sensors altering firing rate and a graphene dopamine sensor producing dose-dependent bidirectional responses. When tested against live cardiomyocytes, the artificial neuron fired in response to a drug that disrupted normal cell rhythm, proving real-time electrical dialogue with living tissue.
A custom mRNA cancer vaccine that caused a tennis ball-sized tumor to shrink by half was created for Rosie, an eight-year-old rescue dog in Sydney, after her owner Paul Conyngham used ChatGPT to navigate a path through experimental cancer treatment. Conyngham, an AI specialist and electrical engineer who co-founded Core Intelligence Technologies, first consulted the chatbot for immunotherapy leads, then had Rosie's DNA sequenced at the University of New South Wales following ChatGPT's suggestion. Associate Professor Smith of UNSW's centre for genomics noted that mapping the tumor's mutations enabled researchers to identify the precise biological drivers of the cancer. Conyngham processed the genomic data through multiple analysis pipelines before connecting with Thordarson, an Icelandic nanomedicine expert at the UNSW RNA Institute, who condensed the findings into a tailored mRNA vaccine formula. Professor Thordarson stated that Rosie's case is the first personalized cancer vaccine developed for a dog, and that it demonstrates how rapidly and effectively mRNA technology can be deployed — with direct implications for human oncology.
A neuroadaptive control system that lets robots respond to human brain signals in real time is being developed by a team at Oklahoma State University, led by researcher Hemanth Manjunatha. The system detects error-related potentials — electrical patterns generated in the brain's anterior cingulate cortex the instant a person senses something is going wrong — before any physical correction can be made. Using a wearable EEG cap, the setup captures those signals and feeds them into a shared-control robotic framework that can slow, stop, or hand back control within milliseconds. Manjunatha emphasized that in high-pressure environments like nuclear decommissioning or underwater inspections, the unpredictability of the setting makes full robot autonomy unsafe. Signal Temporal Logic is used to keep all robot responses within defined behavioral limits, ensuring brain-driven corrections never push the system outside safe operating boundaries.
Cleared by China's National Medical Products Administration, a brain-computer interface system built by Borui Kang Medical Technology in Shanghai has become what regulators claim is the world's first commercially authorized BCI device. The system targets adults aged 18 to 60 with quadriplegia from cervical spinal cord injuries, specifically those who retain some upper arm movement but have lost hand function for at least one year. Borui Kang's device reads brain signals through electrodes implanted outside the dura mater — a minimally invasive approach that avoids deep brain penetration — and wirelessly converts the user's movement intention into commands that drive a robotic glove. China's government has designated BCI as a priority "future industry" in its latest development plan, signaling a national commitment to the sector. The approval positions China as a direct competitor to U.S. neurotechnology firms, including Elon Musk's Neuralink, which is pursuing similar implantable devices for neurological disorders.
A breakthrough experiment was carried out by GuoXing Aerospace Technology and Shanghai Jiao Tong University, marking what is reportedly the first time a humanoid robot was controlled using AI inference conducted entirely from space. A voice command issued on the ground was transmitted to satellites in low Earth orbit, where Alibaba's Qwen3 large language model — running inside a radiation-protected enclosure — processed the instruction and determined the robot's movements. The resulting commands returned to Earth and were interpreted into physical actions by the OpenClaw AI agent. GuoXing Aerospace, based in Chengdu, already has 12 satellites in orbit and plans to expand to 1,000 by 2030 and a full constellation of 2,800 units by 2035. The team says the architecture could support drones, autonomous vehicles, and robotic systems in areas where ground-based networks like 5G and fiber optics are unavailable.
🔊 Industry Insights & Updates
Pioneered by Humboldt University Berlin researchers, a new ultrafast laser technique has overcome one of the key barriers to building a practical quantum internet using diamonds. Doctoral candidate Cem Güney Torun and former research assistant Mustafa Gökçe led the work, introducing the SUPER method to efficiently excite tin vacancy centers in diamond crystals and extract pure, usable single photons. Traditional approaches required complex filtering systems that degraded efficiency and blocked scalability — a problem the SUPER method eliminates by controlling quantum states at speeds previously unavailable to researchers. Critically, the Humboldt team demonstrated that the technique maintains the internal quantum state of the system, which is essential for entangling remote nodes across a quantum network. The study integrated diamond nanofabrication, ultrafast optical technologies, and theoretical modeling into a unified experimental framework.
A five-fingered robotic hand with 20 degrees of freedom has been unveiled by South Korean startup Tesollo, marking a strategic shift from high-specification research hardware toward commercially deployable dexterity for humanoid robots. Named the DG-5F-S and weighing just 880 grams, the hand was engineered by the Incheon-based company to fit within the size and weight constraints that have previously blocked robotic hands from being integrated into real humanoid platforms. Backdrivable joints give the hand mechanical compliance, allowing it to absorb external shocks and interact safely with its environment without sustaining damage. For labs with simpler needs, Tesollo also offers a 15-DoF version with a reduced footprint and lower control complexity. The global five-finger robotic hand market is projected to reach $876 million by 2030, and Tesollo's integration-first approach positions it as a strong commercial contender as companies like Tesla and Figure accelerate development of general-purpose humanoids.
Unveiled at NVIDIA GTC 2026 in San Jose on March 18, the TM Xplore I is a wheeled humanoid robot developed by Taiwan's Techman Robot in collaboration with Quanta Technology and NVIDIA, built for deployment in smart manufacturing environments. Techman designed the robot with a humanoid upper body mounted on a wheeled mobile base — a deliberate choice over bipedal locomotion that prioritizes operational reliability and stability on factory floors. The robot is powered by NVIDIA's Jetson Thor module for edge AI computing, and integrates NVIDIA Isaac, FoundationStereo, and Isaac GR00T to cover simulation, depth perception, and autonomous dexterity. Vision-language capabilities allow the TM Xplore I to interpret both visual and spoken cues, enabling reasoning and decision-making during complex industrial tasks. Techman plans to commercialize the wheeled robot and its associated patents before the end of this year, targeting sectors including semiconductor fabrication, electronics assembly, and automotive manufacturing.
A strategic partnership was signed in Shenzhen on March 16 by Chinese robotics company UBTech and Siemens Digital Industries Software, targeting the mass production of 10,000 humanoid robots annually by 2026. UBTech founder and CEO Zhou Jian confirmed a sharp rise in orders this year, making the leap from prototype to production a defining challenge for the company. Siemens will provide its full industrial software suite — covering product design, simulation, process planning, and manufacturing management — to digitize UBTech's entire production lifecycle. UBTech already recorded over 1.4 billion yuan in humanoid robot orders in 2025, spanning manufacturing and logistics customers, and has begun delivering its Walker S industrial humanoid. Siemens will also provide technical training to help UBTech build the skilled workforce needed for high-volume output.
🧬 BioTech
Turning discarded plastic bottles into Parkinson's medicine was achieved for the first time by the University of Edinburgh, in research led by Professor Stephen Wallace and published in Nature Sustainability. The Edinburgh team's process breaks PET plastic into terephthalic acid — its core carbon building block — before feeding that acid to genetically modified E. coli bacteria that rearrange the molecules into L-DOPA, the dopamine precursor medication most commonly prescribed to Parkinson's patients. Professor Charlotte Deane of UKRI EPSRC praised the work, saying it shows how carbon lost to landfills can be converted into high-value products that enhance lives. Dr. Fletcher of the Industrial Biotechnology Innovation Centre added that by showcasing a harmful material being transformed into something medically beneficial, the team is proving that sustainable biological manufacturing is not theoretical but practically achievable. The researchers believe similar bio-upcycling methods could eventually be extended to produce flavors, fragrances, and industrial dyes from plastic waste.
A bacteria-powered bioelectronic sensor has been developed by scientists at Rice University, capable of detecting specific chemicals in liquids by harnessing the natural electrical signals of living microbes. The device uses a chitosan hydrogel — a soft, porous material derived from crustacean shells — to keep bacteria anchored close to an electrode while allowing liquids to flow through. Redox-active mediators are woven directly into the polymer to ensure stable electron transfer between bacteria and electrode. In a proof-of-concept test, the team engineered L. plantarum bacteria to detect sakacin P, an antimicrobial peptide used as a food preservative in milk, generating a measurable electrical signal within hours. Corresponding author Rafael Verduzco and lead author Xinyuan Zuo say the system opens the door to living bioelectronic devices for sensing, chemical production, and hazardous compound removal across food, wastewater, and industrial settings.
Clinically meaningful recovery from chronic stroke was achieved using an at-home brain-computer interface in the first randomized controlled trial of an FDA-cleared, non-invasive BCI therapy, announced by Kandu, Inc. of Van Nuys, California. The IpsiHand System, which received FDA Breakthrough Device clearance in 2021, works by detecting ipsilateral motor intent signals from the unaffected brain hemisphere and using them to drive hand rehabilitation exercises without any surgical implant. Results presented by Chief Scientific Officer Dr. Eric Leuthardt at the International Stroke Conference 2026 in New Orleans showed that BCI patients gained a mean of 6.0 points on the Upper Extremity Fugl-Meyer scale versus 1.5 points in the control group — a 4.5-point treatment advantage that crossed the threshold for clinical significance. Enrollment in the 62-participant study was halted early after an interim analysis confirmed efficacy, with 55.5% of BCI patients achieving a meaningful functional response compared to just 9.6% in the conventional exercise group. CEO Leo Petrossian said the findings fundamentally challenge the long-held belief that stroke recovery permanently plateaus within the first few months after injury.
💡Products/tools of the week
Paperclip is an open-source, self-hosted orchestration platform that runs autonomous AI-driven companies by hiring, organizing, and coordinating LLM- or agent-based workers into structured org charts. Users can define company goals, assign roles such as CEO, CTO, engineers, and marketers, and connect any runtime that can receive a heartbeat. AI agents execute tasks with goal-aware context, scheduled heartbeats, persistent state, full ticketing, and immutable audit logs, while per-agent budgets and governance keep every action traceable, controllable, and cost-limited. Teams use Paperclip to replace messy agent scripts and dozens of tabs with a scalable, auditable system for automating product development, marketing, support, or entire multi-company operations.
Trio is an AI-powered real-time phone interpreter that connects instantly the moment a service number is dialed, delivering simultaneous translation in just 3–6 seconds with no app or setup required. Built for small businesses navigating multilingual calls, Trio uses AI models optimized for business-grade accuracy, industry-specific terminology, and smooth turn-taking guided by smart audio prompts. The service supports five premium languages — Mandarin, Spanish, Portuguese, Korean, and Japanese — and offers analytics, integrations, and cost-effective plans that cut interpreter costs by up to 70% compared to traditional services. Businesses use Trio to handle multilingual support quickly, reduce overhead, and scale without contracts or special equipment.
InsForge is an AI-native backend platform that gives coding agents everything they need to build, manage, and deploy full-stack applications autonomously, without manual API key setup or infrastructure configuration. The platform bundles Postgres databases, authentication, S3-compatible storage, serverless edge functions, vector search and embeddings, and a Model Gateway that unifies access to OpenAI, Anthropic, Gemini, Grok, and more under one roof. An MCP agent interface lets agents inspect schemas, provision resources, call models, and deploy directly, so tools like Cursor, Claude, and GPT can scaffold and ship backends faster with fewer tokens and fewer errors than traditional BaaS platforms. Developers and teams use InsForge to simplify prototyping, power agentic workflows, and execute one-click deployments at a scale that was previously out of reach.
ShipSafe uses AI-driven orchestration powered by Gemini 2.5 Pro to run and interpret active security scans across industry-standard tools including ZAP, Nmap, Nikto, SSLyze, and Playwright, delivering a fast OWASP Top 10 deep audit without requiring a costly penetration test. The platform automates the entire scanning process end to end and translates raw findings into a plain-English remediation report that founders and builders can act on immediately, regardless of their security background. ShipSafe was built for startups and small teams that need to prove security posture quickly, close enterprise deals that require compliance evidence, and fix vulnerabilities before they become liabilities. Teams get enterprise-grade security validation at a fraction of the cost and time of traditional pen testing.
