Weekly Piece of Future #153
From Synthetic Skin to Throat Sensors and Orbital Cancer Labs
Hey there, fellow future-addicts!
Welcome to this week's edition of Rushing Robotics! Another week, another leap forward in artificial intelligence and emerging technologies. From molecules designed on demand to AI code that now powers nearly a third of all new software, we're watching intelligence reshape how we discover, build, and heal. This edition brings you breakthrough research that's shortening the path from lab bench to real-world impact, industry moves positioning AI for the physical world, biotech innovations that quite literally grow inside living cells, and the tools putting these capabilities into more hands every day.
🤯 Mind-Blowing
PropMolFlow from University of Florida and NYU generates molecular candidates 10 times faster than existing methods without sacrificing accuracy, producing chemically valid structures with correct bonding patterns over 90% of the time . AI-assisted coding now accounts for 29% of all newly written U.S. software functions—up from 5% in 2022—boosting productivity by 3.6% and adding $23-38 billion in annual economic value . Revoice, a washable wearable choker from Cambridge, uses throat sensors and AI to give stroke survivors their voice back by interpreting muscle vibrations in real time . A new analogue chip from Politecnico di Milano could cut computing energy by up to 5,000 times by processing calculations directly in memory . Integrated Quantum Technologies' AIQu™ VEIL™ platform lets businesses deploy AI without exposing sensitive data through dynamically computed anonymity .
🔊 Industry Insights & Updates
Microsoft's Rho-alpha interprets natural language commands for robots doing two-handed tasks, pushing AI into the physical world . The AM2PC project's 3D-printed cooling hit 600 watts—50% above target—while cutting data center energy use . MuWave secured funding to commercialize gyrotron tech for fusion power . TISQ integrates p-type and n-type semiconductors into single molecules for more stable organic solar cells .
🧬 BioTech
Slovenian researchers printed 3D micro-objects inside living cells using ultrafast lasers . Zurich clinicians are using denovoSkin lab-grown grafts to treat severe burns from the January 1 Crans-Montana fire that killed 40 . SPARK Microgravity announced Europe's first commercial orbital cancer lab to study tumors without gravity's interference.
💡 Products/Tools of the Week
AyeWatch scans a billion sources 24/7 with AI for real-time alerts . Gobii deploys AI workers for web automation without APIs . CanvasGPT combines research and full-stack code in one workspace . Janus animates avatars from webcam feeds for streaming .
🎥 Video Section
This week's video lineup shows humanoid robots making the leap from lab demos to real work environments. UBTECH showcases the first humanoid robotic worker deployed at SANY RE, while PNDbotics' Adam full-size humanoid demonstrates increasingly fluid movement with its Charleston Dance Version 2.0 . Skild AI reveals how their systems are learning complex manipulation tasks simply by watching human videos, and Agile Robots SE unveils Agile ONE—their next evolution in physical AI that promises more adaptive, capable robotic systems .
AI is moving from prediction to creation—building molecules, writing software, restoring voices, walking into factories. The pace is breathtaking and the best is still ahead. Stay hungry, stay futurish!
🤯 Mind-Blowing
A new generative AI method developed by researchers at the University of Florida and New York University can produce molecular candidates approximately 10 times faster than current techniques without sacrificing accuracy or chemical validity, offering promise for accelerated drug and materials discovery. The PropMolFlow system inverts traditional scientific discovery by allowing scientists to specify desired properties and then find matching molecular structures, rather than discovering useful compounds by accident and later understanding why they work. According to Mingjie Liu, an assistant professor in the University of Florida's Department of Chemistry and one of the paper's authors published in Nature Computational Science, computational speed directly translates to discovery speed in this field, making the advance particularly meaningful. The innovative algorithm finds more direct paths from random noise to valid molecular structures, consistently outperforming baseline models on structural validity by generating molecules with correct bonding patterns and appropriate geometries more than 90 percent of the time.
A new study published in Science reveals that AI-assisted coding has reached 29% of all newly written software functions in the United States by early 2025, up from just 5% in 2022, according to research from the Complexity Science Hub that analyzed over 30 million Python contributions from approximately 160,000 developers on GitHub. The research team led by Simone Daniotti used a specially trained AI model to identify AI-generated code through tools like ChatGPT or GitHub Copilot, finding that generative AI increased programmers' productivity by 3.6%, translating to between $23 and $38 billion in annual value for the U.S. economy. Frank Neffke, who leads the Transforming Economies group at CSH, noted that while the U.S. leads at 29%, Germany reaches 23%, France 24%, and India 20%, with China at 12%. Less experienced programmers use generative AI in 37% of their code compared to just 27% for experienced programmers, but the productivity gains are driven exclusively by experienced users
Regained speech after stroke is the goal of Revoice, a comfortable, washable wearable choker that uses sensitive sensors and AI to interpret speech-related throat signals. The University of Cambridge team designed it for people with dysarthria, where weakened facial, oral, and throat muscles make speech difficult even when the person knows what they want to say. The system avoids invasive brain implants by capturing throat muscle vibrations and heart rate, then generating intended words and sentences in real time.
Reduced by up to 5,000 times, energy consumption in computing could drop dramatically with a new analogue chip that executes calculations directly within memory arrays. Daniele Ielmini, a professor in the Department of Electronics, Information, and Bioengineering at Politecnico di Milano, led the development of the fully integrated analogue accelerator built using standard CMOS technology compatible with current silicon manufacturing. Testing revealed the chip achieved accuracy comparable to traditional digital systems but with reduced power consumption, lower computing latency, and a smaller silicon footprint.
A newly introduced AI infrastructure platform enables businesses to implement artificial intelligence without exposing sensitive data, addressing persistent privacy and security challenges for organizations under stringent regulations. Integrated Quantum Technologies has unveiled its AIQu™ VEIL™ platform, which processes data in anonymized and compressed formats before integration into AI systems using a technique called Dynamically Computed Anonymity or DCA™ that transforms data into vectorized forms. Guib, CEO and Chairman of Integrated Quantum Technologies, stated that VEIL™ enables companies to harness AI without compromising their most critical data. Jeremy Samuelson, Executive Vice President of AI and Innovation who conceptualized VEIL™, noted that by processing exclusively mathematically anonymized data, they are creating effective and secure AI systems at scale.
🔊 Industry Insights & Updates
Launched by Microsoft Research to extend robot capabilities beyond controlled factory settings, Rho-alpha is a new AI model designed to tackle a persistent robotics challenge: adapting to unpredictable, real-world conditions. The system represents Microsoft's first robotics-centric model from its Phi vision-language AI lineage and is positioned as part of a broader transition toward "physical AI," where intelligent agents engage directly with the tangible world rather than being limited to digital realms. Unlike conventional industrial robots that rely on inflexible task scripts, Rho-alpha interprets natural language commands and converts them into control signals for robots executing intricate two-handed manipulation tasks, with Microsoft currently assessing the system on dual-arm platforms and humanoid robots.
A new passive two-phase cooling approach has been demonstrated to help data centers handle rising server and GPU heat loads while cutting the energy used for cooling. The AM2PC project, led by the Danish Technological Institute and Heatflow with partners from Belgium and Germany, developed a 3D-printed aluminum cooling component intended for data centers and HPC environments. In trials, the solution reached 600 watts of performance, exceeding the original 400-watt target by 50%.
Backed by seed investment intended to accelerate deep-tech commercialization, MuWave will build and refine next-generation gyrotrons and related microwave products that can reduce risk and speed progress toward practical fusion power. Co-founder Helen Webster called the initial funding a milestone that validates the problem the team is solving and supports building momentum toward larger impact, while the UKI2S investment—managed by Future Planet Capital—highlighted the goal of turning scientific know-how into long-term industrial and economic outcomes beyond fusion alone. MuWave’s near-term plan includes producing and supplying high-powered microwave products within the UK while developing an export-ready technology base for global markets.
Formed as a “shortcut” toward better organic thin-film solar cells, TISQ addresses a key bottleneck: the interface between p-type and n-type materials must be precisely tuned to separate and transport charges quickly after light absorption. Osaka Metropolitan University’s Takeshi Maeda said traditional fabrication often physically mixes p-type and n-type molecules, but small processing changes can cause inconsistent mixing, unstable structures, and reduced device performance. The team instead pursued a single-molecule strategy where p- and n-type semiconductor components are integrated into one molecule so nanoscale p/n heterojunctions can arise through molecular self-assembly.
🧬 BioTech
Printed 3D micro-objects were formed inside living cells, demonstrating a way to fabricate free-standing structures where cells normally cannot take up large solids. The Slovenia-based research team introduced 10–15 micrometer droplets of a commercial resist into HeLa cells and then scanned an ultrafast laser to harden the material only where needed, producing precise internal microstructures. The approach points toward new experiments that monitor cellular activity or test how cells respond to physical stimuli, while keeping the surrounding environment largely intact.
Began after a catastrophic New Year’s fire in Crans-Montana, Switzerland, an advanced treatment effort is now using lab-grown skin grafts to help survivors with severe burns recover. The blaze at the Le Constellation bar on January 1, 2026 killed 40 people and injured 119, with an estimated 80–100 survivors suffering serious burns, many over more than 60% of their bodies. Clinicians in Zurich are using denovoSkin, a living skin graft grown from each patient’s own cells, to support wound closure and healing in the most severe cases.
Announced at a public forum titled "Curing Cancer in Space," the orbital laboratory will enable experiments on three-dimensional tumor growth and cell signaling without gravitational interference. SPARK Microgravity executives explained that gravity influences cell growth, positioning, and interactions on Earth, creating limitations that are lifted in orbit where researchers can investigate intricate cancer behaviors more clearly. CEO and co-founder Allison Bajet illustrated the difference using an audio analogy, comparing Earth-based experiments to listening to a symphony in a construction zone while microgravity is like entering a soundproof room where every biological note becomes crystal clear.
💡Products/tools of the week
New always-on monitoring is now available with AyeWatch, a 24/7 AI-powered tool from AyeSoul that uses its SOUL Watcher multi-agent AI to scan over a billion sources across news, blogs, social, research papers, and webpages. AyeSoul’s AyeWatch sends real-time, rule-based alerts only when relevant matches or significant changes occur, filtering duplicates and noise while learning from user feedback to improve relevance. By prioritizing actionable items, AyeWatch helps researchers, investors, journalists, founders, and other power users stay ahead without constant manual searching, supported by cross-platform apps and scalable free-to-paid plans.
Always-on web automation is now easier with Gobii, a cloud platform that deploys AI-powered digital workers to run web-based tasks and continuous monitoring 24/7. Gobii’s agents can browse sites even without APIs, execute JavaScript, rotate proxies, extract and synthesize data, and take action on scheduled cadences or event triggers. With prebuilt worker templates for recruiting, sales, intel, and compliance plus integrations with Slack, Google Drive, CRMs, and APIs, Gobii helps teams offload repetitive research, enrichment, and operational workflows so AI handles data collection and basic decision-making while humans focus on higher-value work.
Live product building is now centralized in CanvasGPT, an AI-powered infinite workspace that uses autonomous canvas agents like Patch plus integrated models such as Gemini, Claude Opus, Nano Banana Pro, ElevenLabs, and Kling to turn ideas into working products. CanvasGPT supports brainstorming, web research, high-fidelity image/audio/video generation, full-stack code production, and workflow automation directly on a persistent, shareable canvas. With AI embedded into core components like a built-in database, auth, and real-time sync, CanvasGPT helps non-developers and teams prototype, build, and deploy interactive apps without tool handoffs, while a credits-based system scales usage and model choice for faster creation and fewer silos.
Real-time avatar animation is now available with Janus, an AI-powered character animation platform that turns a live webcam feed into animated avatars for streaming, video calls, and content creation. Janus uses AI facial-tracking and animation models to map expressions and movements onto pre-made or custom characters instantly, running in-browser or through a Chrome extension with no special hardware required. With broad streaming support, Janus enables live output to Discord, Zoom, Twitch, YouTube, OBS, and more while keeping setup lightweight for creators and teams.
It's interesting how you consistently highlight the breathtaking pace of AI development, especially as I see coding assistants becaming integral for student projects.