Weekly Piece of Future #138
From Biological Computers to Artificial Neurons and Quantum Sensing
Hey there, fellow future-addicts!
Welcome to this week's edition of Rushing Robotics! Every week, science and technology take leaps that feel like they belong to tomorrow. From lab-grown brains that can learn to play Pong to exosuits built with sewing threads, the pace of discovery is accelerating, reshaping not only what machines can do but what life itself can become.
🤯 Mind-Blowing
Frontier science is rewriting the rules of what intelligence and biology can be. Imagine a shoebox-sized brain made of human neurons wired into silicon, trained to process information like a living computer. Or artificial neurons engineered in the lab, running at a fraction of the energy of today’s AI hardware. Add in chip-scale lasers powering quantum labs and even a handheld “bone-healing gun” that prints scaffolds directly onto fractures, and the line between science fiction and science fact is vanishing fast.
🔊 Industry Insights & Updates
The robotics and AI industries aren’t standing still—they’re evolving at breakneck speed. NVIDIA is giving humanoid robots both sharper reasoning and stronger physicality, uniting physics and intelligence in a way that makes them more capable in human environments. Meanwhile, memory research promises to cut energy demands by a factor of ten, new textile-based exosuits are delivering VR feedback for under $500, and autonomous robots are learning to forget irrelevant information just like humans do. Even China is building “robot boot camps” where machines can train in simulated real-world environments at scale.
🧬 BioTech
Biology itself is being hacked, rebuilt, and reimagined. Researchers are converting skin cells into eggs that can develop into early human embryos, hinting at a new future for fertility treatments. At the same time, engineers are building “AggreBots” from lung cells—tiny programmable machines powered by living cilia. And in surgical training, 3D-printed organs filled with blood-like fluids are proving more realistic than ever, offering safer, smarter ways to prepare the next generation of doctors.
💡 Products/Tools of the Week
The future is also arriving on your desktop. WorkBeaver watches your screen, learns your repetitive tasks, and automates them without code. GrainODM uses AI to grade grain in seconds with near-perfect accuracy. myNeutron gives you a digital memory that unifies your scattered knowledge across email, cloud drives, and the web. And Scale’s SEAL Leaderboards are setting a new bar for evaluating which AI models perform best in the real world—not just in benchmarks.
🎥 Video Section
Some breakthroughs are better seen than described. This week’s picks include a soft, low-cost exosuit delivering full-body haptics, DeepMind’s new generative AI design workflows, and a humanoid robot ready for mass production.
Brains in boxes, robots in gyms, bones healing with “guns”—the headlines of tomorrow are already wild today. But what sounds like sci-fi quickly becomes everyday tech. Stick with us, and you’ll always be one step ahead of the future sneaking into the present. Stay hungry, stay futurish!
🤯 Mind-Blowing
A lab-grown brain made of 800,000 human neurons that can play Pong has been unveiled as the world’s first code-deployable biological computer. Known as CL1, the shoebox-sized system integrates living neurons with silicon chips to form a synthetic biological intelligence, combining the adaptability of human cells with the processing stability of silicon. The project builds on DishBrain, an earlier platform from Australian startup Cortical Labs that connected neurons to a microelectrode array, allowing them to interact with a Pong simulation in real time. CL1 advances this work by embedding human neurons into a chip interface where they are sustained in a nutrient solution and trained through electrical signals to perform tasks. According to Cortical Labs CEO Brett Kagan, the system allows labs to directly study how neurons process information—a first for biological computing. Each unit will cost about USD 35,000, consume 850–1,000 watts, and operate without an external computer. Potential uses include disease modeling, drug testing, adaptive robotics, and pharmaceutical research, though its neurons typically expire within six months.
An artificial neuron created in the lab now functions almost like one found in the human body, demonstrating electrical properties that closely mirror those of biological neurons. The breakthrough builds on earlier work using protein nanowires produced by electricity-generating bacteria. Researchers believe this discovery could pave the way for ultra-efficient computers modeled after biology and even systems capable of directly interfacing with living cells. The human brain operates on about 20 watts of power, while training or running a large AI model can demand more than a megawatt for comparable tasks. Closing this efficiency gap is the ultimate goal of the new neuron design. Because it runs at low voltage, the artificial neuron could be integrated not only into future bio-inspired computing architectures but also into medical devices designed to communicate directly with biological tissue.
Scientists at the University of California, Santa Barbara, have developed ultra-stable, low-noise lasers integrated onto silicon nitride chips, opening new horizons in quantum sensing and computing. Traditionally, such laser systems required bulky optical tables, but the team is working to shrink them into palm-sized chips. The breakthrough could revolutionize fields ranging from navigation and climate monitoring to timekeeping and quantum technologies. With lasers, beam delivery, and optical control now operating on-chip, the researchers are approaching the creation of a complete optical laboratory in miniature form. They have also extended their platform to trapped ions—one of the leading architectures for quantum computers—demonstrating the ability to generate ion-based qubits using integrated lasers. Achieved in partnership with the University of Massachusetts Amherst, this marks the first time ion qubits have been created directly with chip-scale lasers, bringing compact quantum processors and sensors within reach.
Researchers at Sungkyunkwan University in Korea are developing a novel “bone-healing gun” that could transform fracture treatment by making it faster and less invasive. The device works like a handheld 3D printer, but instead of printing plastic, it extrudes biodegradable polymer scaffolds directly onto broken bones during surgery. The tool melts polymer “bullets” at just 60 °C—low enough to avoid damaging nearby tissue—while forming a patient-specific framework to guide bone regeneration. The researchers used polycaprolactone, an FDA-approved thermoplastic that naturally degrades in the body over several months, and combined it with hydroxyapatite, a mineral that supports new bone growth. By carefully adjusting the mixture, they created a material that melts safely at 60 °C, bonds securely to bone, maintains strength through the healing period, and then gradually disappears as natural bone tissue takes over.
🔊 Industry Insights & Updates
Humanoid robots are gaining smarter muscles and sharper intelligence with NVIDIA’s latest innovations, as the Newton Physics Engine and Isaac GR00T N1.6 combine to bring safer, more capable humanoids by merging physics, reasoning, and AI infrastructure. In effect, robots just received both a new brain and a stronger body. NVIDIA has rolled out a broad set of updates designed to accelerate humanoid robotics and physical AI. The company introduced the open-source Newton Physics Engine, the Isaac GR00T N1.6 foundation model for robots, and enhanced AI infrastructure, all tightly integrated into its Isaac Lab platform. Together, these elements provide researchers and developers with a unified, open, and accelerated robotics stack. The tools aim to accelerate iteration, standardize testing, unify training with on-robot inference, and ensure safer transfer of skills from simulation to real-world applications. These updates also highlight NVIDIA’s drive to establish the defining platform for general-purpose robotics—particularly humanoids, which require sophisticated reasoning and precise physics to operate effectively in human-centered environments.
Researchers at Chalmers University of Technology in Sweden have announced a breakthrough that could transform digital memory technology. The team has engineered an atomically thin material capable of allowing two opposing magnetic forces to coexist, which reduces the energy required by memory devices by a factor of ten. Memory is central to nearly all modern technologies, powering AI systems, smartphones, computers, self-driving cars, home appliances, and medical equipment. Magnetism plays a critical role in digital storage, as manipulating electron behavior with magnetic fields and electric currents enables faster, smaller, and more efficient chips. However, the rapid rise in global data has pushed conventional memory toward its limits, with projections showing that within decades, digital memory could consume nearly 30 percent of the world’s energy. This has accelerated the need for alternatives that are both energy-efficient and scalable. The Chalmers team is the first to demonstrate that a layered, two-dimensional material can unite two distinct magnetic forces, cutting the energy demand of memory devices by tenfold.
A new lightweight exoskeleton suit, known as Kinethreads, is delivering lifelike VR feedback using nothing more than threads, compact motors, and a Raspberry Pi. Unlike traditional full-body exoskeletons, which often cost six figures, Kinethreads can be built for under $500 with off-the-shelf components and clever textile engineering. The suit wraps snugly around the chest, arms, and legs like a second skin. Nylon threads threaded through fabric channels connect to small motors, functioning like artificial tendons. When triggered, the threads tighten and contract, pulling directly on the wearer’s muscles to guide movement and stabilize joints. Control comes from a Raspberry Pi running simple Python scripts, allowing for customizable scenarios. In tests, the system convincingly replicated experiences such as lifting virtual objects or bracing against impacts. Kinethreads shows that realistic haptic feedback and muscle assistance no longer require bulky, prohibitively expensive exoskeletons.
New “Physical AI” technology has been shown to improve the navigation of autonomous mobile robots in logistics centers and smart factories by enabling them to forget irrelevant data while focusing on critical, real-time information. Developed at South Korea’s Daegu Gyeongbuk Institute of Science and Technology (DGIST), the system is modeled on the human-like process of “spreading and forgetting” social issues. By adopting this principle, robots are able to distinguish between immediate, important obstacles and outdated, unnecessary information, streamlining their movement. “We have mimicked the social principle of forgetting unnecessary information while retaining only important information to enable efficient movement. This study is significant in that it shows how Physical AI is evolving to resemble human behavior,” explained Professor Kyung-Joon Park, the lead researcher. Testing revealed an average driving time reduction of up to 30.1% and task throughput gains of up to 18.0%. The researchers say such improvements could directly increase capacity and revenue for large-scale logistics and manufacturing operations. They emphasize that this advance demonstrates how robots are shifting beyond simple obstacle-avoidance toward becoming Physical AI systems that incorporate social principles to operate more autonomously.
China is preparing to launch a network of robot “boot camps” in several major cities, including hubs like Beijing and Shanghai, to accelerate the development of its humanoid robotics sector. These facilities will function as gyms or obstacle courses where robots can practice in simulated real-world environments. The largest of the centers will be built in Beijing’s Shijingshan district, covering more than 108,000 square feet (10,000 square meters) and producing over 6 million data points annually. Within these spaces, robots will train in scenarios such as factories, retail stores, eldercare facilities, and smart homes, generating standardized datasets that robotics companies can use to enhance their AI systems. The initiative coincides with China’s broader strategy of rapidly expanding robotics adoption across multiple industries nationwide.
🧬 BioTech
Scientists have demonstrated a proof of concept that converts skin cells into eggs capable of developing into early human embryos. The approach, known as in vitro gametogenesis, could eventually provide new options for treating infertility. Unlike previous methods that reprogram stem cells into gametes, this team employed somatic cell nuclear transfer—the same technique used in cloning Dolly the sheep in 1997. In their study, researchers created 82 functional oocytes and fertilized them with sperm. Most embryos stopped developing at the 4- to 8-cell stage due to chromosomal abnormalities, but around 9 percent progressed to the blastocyst stage after six days. The process involved transplanting the nucleus of a skin cell into an enucleated donor egg, whose cytoplasm induced the skin cell nucleus to discard half of its chromosomes, mimicking meiosis. The resulting haploid eggs could then be fertilized through IVF, producing embryos with equal genetic contributions from both parents.
At Carnegie Mellon University, engineers have pioneered a new method for creating “designer” biological robots from human lung cells. These microscale machines, called AggreBots, represent a new class of biobots—living, man-made systems capable of autonomous movement and programmed behaviors. While most biobots rely on muscle fibers to power contraction and relaxation, the CMU team is exploring cilia, the tiny hair-like structures that propel fluids in the human body and help single-celled organisms like Paramecium swim. Designing cilia-powered robots, or CiliaBots, has long been hindered by difficulties in shaping and organizing cilia structures with precision. To overcome this, the Ren lab developed a modular assembly strategy in which tissue spheroids derived from lung stem cells are spatially aggregated to form AggreBots with controllable motility. This system even allows for the inclusion of spheroids with genetic mutations that silence activity in specific cilia regions, offering a way to tune movement patterns with unprecedented precision.
A new form of 3D-printed tissue containing blood-like fluids has been developed to mimic real human organs for surgical training, and surgeons who tested the replicas rated them superior to conventional practice models, fueling optimism for safer and more effective medical training. Earlier 3D-printed tissues failed to capture the complexity of natural organs. Researchers at Minnesota devised a method to control the shape and size of microscopic patterns within the printed material. These patterns determine the strength and elasticity of the tissue, providing it with lifelike mechanical behavior. The team also designed a mathematical formula to predict how the tissues respond under stress. To make the models even more realistic, they infused blood-like fluids during printing, with microcapsules encasing the liquid to prevent evaporation or disruption of the printing process.
💡Products/tools of the week
WorkBeaver is an AI-driven automation solution designed to remove repetitive tasks from your daily computer work by observing and learning directly from the way you operate. Unlike conventional automation tools that demand coding skills or intricate configuration, WorkBeaver enables users to simply carry out a process once in the normal way, after which the AI automatically repeats those exact steps in future sessions. With artificial intelligence at its core, the system adapts to interface changes and operates seamlessly across any visible application on your screen, whether browser-based or desktop software. By interpreting context and adjusting intelligently to interface variations, WorkBeaver empowers individuals, small businesses, and large organizations to reclaim valuable time otherwise lost to manual tasks such as data entry, form completion, and email follow-ups—all without technical expertise or risks to data security, as all processing occurs locally on the user’s machine.
GrainODM is an advanced AI-powered solution that revolutionizes grain analysis by replacing slow, inconsistent manual inspections with rapid and highly reliable automated testing. While traditional inspections may take up to 30 minutes and often fail to detect certain defects, GrainODM delivers precise results in just three seconds with an impressive 99.8% accuracy rate. The platform enables mills, processors, and food manufacturers to achieve standardized grading, minimize disputes, and strengthen trust by eliminating human subjectivity. Its design allows operations to scale efficiently without dependence on specialized inspectors, making it ideal for modern agricultural environments. By offering real-time quality assessment and instant reporting, GrainODM redefines how grain quality is measured and validated throughout the industry.
myNeutron is a digital memory system that captures, organizes, and makes information instantly retrievable from scattered sources such as Gmail, Google Drive, and the web. Instead of endlessly re-searching or re-explaining the same material to AI, users can save content with a single click, which the platform then converts into structured “seeds.” These seeds can be queried directly with AI or seamlessly inserted into tools like ChatGPT, Claude, or Gemini for context-rich conversations. With privacy and security built into its foundation, as well as one-click export, myNeutron helps eliminate the frustration of tracking down files, notes, or emails. It transforms fragmented knowledge into a unified, accessible memory layer that saves time and boosts productivity.
Scale SEAL Leaderboards is an AI-driven evaluation platform that delivers trusted rankings of leading large language models such as GPT, Claude, and Gemini by combining carefully curated private datasets with real-world performance metrics. At its core, the platform applies advanced AI to objectively measure models across a wide range of domains, including programming, mathematics, instruction following, and multilingual proficiency, using rigorous AI-powered evaluation frameworks. A unique feature of the system is SEAL Showdown, which employs AI to analyze millions of real user conversations and preferences from diverse demographics around the globe, producing insights segmented by audience type. By relying on Scale SEAL Leaderboards, researchers, enterprises, and developers can confidently determine which AI models align best with their needs, knowing the results are secure from manipulation, comprehensive in scope, and grounded in authentic performance rather than artificially tuned benchmark scores.
