Weekly Piece of Future #154
From AI Creativity Tests to Robot App Stores and Topical Gene Therapy
Hey there, fellow future-addicts!
Welcome to this week's edition of Rushing Robotics! Every week the boundary between "emerging technology" and "working infrastructure" gets a little blurrier. This week brought AI that scores like a creative human, robots running cross-platform apps like smartphones, and gene therapies you can apply like lotion. We're not talking about future promises anymore—we're tracking deployments, benchmarks, and real products hitting real users. Here's what landed this week.
🤯 Mind-Blowing
This week’s breakthroughs blur the line between imagination and implementation. AI systems are now matching or beating average human scores on creativity benchmarks—yet still shine brightest when paired with human guidance. Robots are getting their own cross‑platform “app store,” turning hardware into a software‑defined ecosystem. Materials science and 3D printing are quietly reinventing what’s physically possible, while NVIDIA’s video‑world models let robots plan over long horizons using pure vision. Put together, it’s a picture of tools that don’t just execute instructions, but help us explore entirely new design spaces.
🔊 Industry Insights & Updates
On the industry side, the center of gravity keeps shifting from cars and chips to robots and inference. Tesla is sunsetting its legacy flagships to free capacity for Optimus humanoids—an explicit bet that the next “vehicle” of value is robotic labor, not EVs. Microsoft’s Maia 200 shows how seriously hyperscalers take AI economics, squeezing more tokens per watt and dollar to feed ever-larger models. NASA’s new Athena supercomputer slots AI right alongside rockets and aircraft as core infrastructure, while Hyundai’s humanoid trials reveal how quickly factories may move from fixed automation to adaptable robotic co‑workers.
🧬 BioTech
Biotech is quietly having its own AI‑like moment. A cancer‑linked molecule that clears misfolded brain proteins hints at drug strategies that could decouple Alzheimer’s risk from aging. Meanwhile, CRISPR is stepping out of the lab and onto the skin, with topical gene editing that actually restores function in inherited skin disorders. And by targeting the right exon in a single resistance gene, researchers are reopening treatment paths for head and neck cancers once thought out of options. The pattern: more precise edits, closer to root causes, with delivery methods that feel less like sci‑fi and more like skincare.
💡 Products/Tools of the Week
The tools arriving this week don't just automate—they compress entire workflows into seconds. Mowgli turns napkin sketches into production specs and themed UI, letting designers skip straight to iteration. ModelMonkey brings Claude into your spreadsheet to build financial models, pull live data, and explain every formula it writes. ArchRender bypasses complex 3D lighting setups, rendering photorealistic architectural visuals from sketches or models in moments. And HouseGPTs lets homeowners and agents visualize redesigns and staging instantly, cutting weeks off renovation planning.
🎥 Video Section
If you want to see the future instead of just reading about it, this week’s videos are a good start. Figure’s Helix 02, LimX Dynamics’ autonomous deployments, and Fauna’s real‑world humanoid platform all show different answers to the same question: what does a general‑purpose robot actually look like when it leaves the lab? Watch the gaits, the recovery, the manipulation—then imagine what happens when software updates hit at internet speed.
We’re still early in all of this, which is exactly why it’s so exciting. Creativity benchmarks remind us humans still dominate the top tier, even as AI stretches what’s possible for everyone else. Robots are moving from demos to deployment, and gene editing is edging toward everyday therapies.. Stay hungry, stay futurish!
🤯 Mind-Blowing
Leading AI models have demonstrated creativity levels surpassing average humans in a landmark study comparing 100,000 participants with generative AI systems. Researchers at Université de Montréal employed the Divergent Association Task, a psychology tool measuring creativity by evaluating how unrelated generated words are from each other. The team tested ChatGPT, Claude, and Gemini against human participants, finding that while AI exceeded many humans in divergent creativity scores, half of the participants performed better than the machines, with the top 10% significantly outscoring AI. Professor Karim Jerbi noted that the study highlights how AI profoundly transforms how creators imagine and explore rather than replacing them. Published in Scientific Reports, the research also examined creative writing tasks, revealing that AI expressed maximum creativity when guided by humans, suggesting technology's role as an assistive tool in creative endeavors.
A robot app store went live on Apple's App Store on January 27, 2026, when OpenMind launched its cross-platform marketplace. The platform brings together eight competing companies including UBTECH, Zhiyuan Robotics, and Fourier, signaling the industry's shift from hardware competition to software ecosystem development. OpenMind's OM1 operating system enables developers to write code that works across different robot types, from bipedal humanoids to four-legged robotic dogs. The system uses capability-oriented programming where developers issue commands to abstract motion capabilities rather than specific joints. Security sandboxes prevent direct access to motor drivers, with all motion commands passing strict feasibility checks. The platform includes a web-based simulator and crowdsourced testing network using real robots. OpenMind introduced a skills economy with encrypted execution environments for protection and dynamic pricing based on performance metrics.
A method to transform MXenes into one-dimensional forms that could enhance batteries, biosensors, and wearable technology has been developed by scientists at Drexel University. The scalable technique rolls flat two-dimensional MXene sheets into tubular structures called MXene nanoscrolls, approximately 100 times slimmer than human hair with improved electrical conductivity. Yury Gogotsi explained that one-dimensional forms provide distinct advantages in scenarios requiring rapid transport or mechanical support. Cheng Zhang noted that transitioning from 2D nanosheets to 1D scrolls mitigates nano-confinement issues that challenge ion movement. The process uses water to adjust surface chemistry, creating structural imbalance that causes layers to separate and curl into compact scrolls. The team applied this technique to six distinct MXenes producing grams of nanoscrolls with controlled shapes. The tubular design increases active surface area, enhancing accessibility for ions and molecules crucial for battery performance and chemical sensing.
A novel 3D printing technique named CRAFT has been introduced by researchers from The University of Texas at Austin in collaboration with Sandia National Laboratories, enabling production of objects with diverse mechanical and optical characteristics using a single cost-effective material. The team succeeded in creating a lifelike replica of a human hand by transforming an affordable liquid into a versatile material. Zak Page remarked that they control molecular order in three-dimensional space, completely changing the mechanical and optical properties of a material. CRAFT converts cyclooctene into solid structures through projection of different light patterns, with researchers adjusting light intensity to influence molecular order as material hardens. The team can create parts that are hard and transparent while keeping adjacent sections soft and opaque within a single component. CRAFT presents a superior substitute for medical cadavers by replicating intricate relationships of bone, ligament, and muscle within a single model. Beyond medical applications, CRAFT can produce bioinspired materials for energy absorption in helmets and armor as well as soundproofing.
An innovative method for robot management has been unveiled by NVIDIA as Cosmos Policy, utilizing extensive video prediction models to enhance control and planning functions. NVIDIA refines a pre-existing video world model termed Cosmos Predict using data from robot demonstrations, which has already acquired understanding of how physical environments change over time. Robot actions, physical conditions, and task results are integrated into the model's internal temporal framework, enabling it to forecast subsequent actions and anticipated consequences. Benchmark evaluations show impressive success rates on multi-step tasks requiring long-term reasoning, matching or surpassing existing techniques while needing considerably fewer training demonstrations. The model can produce and assess various potential action sequences during inference, allowing robots to choose actions likely to succeed over extended periods. Hands-on experiments involving dual-arm manipulation successfully completed long-horizon tasks based solely on visual input.
🔊 Industry Insights & Updates
Tesla will cease production of the Model S sedan and Model X SUV as CEO Elon Musk announced during a call with investors on Wednesday, signaling a shift away from traditional automotive programs towards artificial intelligence and robotics. This announcement coincided with the release of the company's latest quarterly earnings revealing a decline in vehicle sales but profits that exceeded expectations. Musk stated that he expects to wind down S and X production this quarter. The Model S debuted over ten years ago cementing the company's reputation in the U.S. automotive market, while the Model X was introduced with unique Falcon-wing doors and high-end appeal. The Fremont, California factory will be repurposed to produce the humanoid robot Optimus, with Musk's long-term vision to manufacture 1 million robots annually at the site
A breakthrough inference accelerator from Microsoft called Maia 200 has been introduced to dramatically improve the economics of AI token generation. The chip, manufactured by Taiwan Semiconductor Manufacturing Co. using the 3nm process, packs 140 billion transistors and is tailored for large-scale AI workloads. Microsoft designed Maia 200 to deliver over 10 petaFLOPS in FP4 and over 5 petaFLOPS in FP8 performance within a 750W SoC power envelope, with its memory subsystem featuring 216GB HBM3e at 7 TB/s and 272MB of on-chip SRAM to keep massive models fed and highly utilized. The new accelerator achieves 30% better performance per dollar than the latest generation hardware in Microsoft's fleet and outperforms competing chips with three times the FP4 performance of Amazon's Trainium3 and FP8 performance above Google's TPU v7. Microsoft is deploying Maia 200 as part of its heterogeneous AI infrastructure to serve the latest GPT-5.2 models from OpenAI and power Microsoft Foundry and Microsoft 365 Copilot.
NASA's most advanced supercomputer named Athena has been unveiled on January 27, intended to support upcoming initiatives in space exploration, aeronautics, and scientific inquiry. Athena stands as the latest flagship within NASA's High-End Computing initiative at Ames Research Center in Silicon Valley. Dan Murphy remarked that exploration has perpetually pushed NASA to the limits of computational capabilities. Athena provides processing power for simulating complex rocket launches and modeling fuel-efficient aircraft designs, enabling NASA to conserve millions by minimizing physical testing. The system acts as a training hub for large-scale artificial intelligence foundational models analyzing vast amounts of satellite and mission data. Athena achieves peak performance exceeding 20 petaflops with 1,024 nodes equipped with AMD EPYC processors and 786 TB of memory.
Humanoid robot trials have begun at Hyundai Motor’s US manufacturing facilities, with proof-of-concept testing underway at the company’s Georgia plant since late last year to evaluate real-world functionality and gather operational data ahead of potential commercialization. The automaker announced plans in early January to implement a scalable platform capable of deploying up to 30,000 humanoid robots annually by 2028 to automate repetitive factory tasks. Hyundai intends to deploy Atlas, the humanoid robot developed by its robotics division Boston Dynamics, at the Georgia site starting in 2028, with initial responsibilities focused on parts sorting before advancing to more complex assembly tasks by approximately 2030. The production-ready Atlas robot, showcased at CES 2026 in Las Vegas on January 5, can lift weights up to 110 pounds and operate in industrial conditions with temperatures ranging from -4°F to 113°F, featuring 56 degrees of freedom to handle high-risk and repetitive activities. Hyundai is developing Software-Defined Factories that gather extensive data to allow robots to continuously refine their skills and is creating Robot Manufacturing Application Centers where robots will be trained in tasks like welding, turning, and assembling, in collaboration with NVIDIA for AI computing and simulation technologies.
🧬 BioTech
A specific molecule found in some cancer cells has been discovered to break down misfolded proteins in the brain, potentially explaining why cancer and Alzheimer's disease rarely occur together. University of Toronto researchers identified that certain cancers release a protein called cystatin-C (Cyst-C), which travels through the body and helps clear toxic amyloid clumps linked to Alzheimer's. The 15-year study used mouse models transplanted with human lung, prostate, and colon tumors, finding that mice with cancer remained free of cognitive decline plaques. Cyst-C binds to harmful amyloid clusters and activates the brain's immune cells, microglia, to break down the plaques. When researchers disabled the Cyst-C pathway, the protective effect vanished. The findings, published in Cell, open doors for developing drugs that degrade misfolded plaques without requiring cancer, and advance the search for existing cancer treatments beneficial for the brain.
A topical CRISPR-based gene therapy has been developed that corrects genetic skin disorders through direct application to the skin. UBC Medicine researchers, working with the Berlin Institute of Health at Charité, created the first gene therapy capable of correcting faulty genes when applied directly to human skin, as detailed in Cell Stem Cell. Dr. Sarah Hedtrich, an associate professor at UBC's School of Biomedical Engineering and senior author, explained that the treatment corrects disease-causing mutations using a safe, scalable approach that addresses the root cause. The therapy demonstrated effectiveness in correcting the genetic mutation behind autosomal recessive congenital ichthyosis, a rare inherited disorder affecting one in 100,000 people. Testing showed the treatment can restore up to 30 percent of normal skin function. The treatment uses lipid nanoparticle technology to deliver gene-editing tools through microscopic openings created by laser, reaching skin stem cells beneath the surface.
An important advance in treating head and neck cancers has been made through CRISPR gene editing that overcomes drug resistance in tumors that stopped responding to treatment. The ChristianaCare Gene Editing Institute team published their findings in Molecular Therapy Oncology, demonstrating a method to restore chemotherapy effectiveness by targeting the NRF2 gene rather than single proteins common in traditional drug development. Head and neck cancer is the seventh most common cancer worldwide with cases expected to rise 30 percent every year by 2030, and many patients reach a point where treatment no longer works. The researchers showed CRISPR can successfully disrupt NRF2 in head and neck cancer cells and esophageal cancer cells, building on earlier lung cancer studies where blocking NRF2 improved survival in animal models. Dr. Natalia Rivera-Torres emphasized breaking through the wall of drug resistance that so many patients face. The study revealed that the location of the CRISPR cut within the NRF2 gene makes a big difference, with targeting exon 4 reducing NRF2 levels by 90 percent compared to less effective exon 2 editing.
💡Products/tools of the week
A new AI design copilot transforms napkin sketches into production-ready specifications and high-fidelity screens. Mowgli guides users through a structured questionnaire, automatically generating user journeys, data models, and themed UI elements. The platform supports natural-language iteration and exports work as Figma files, React+Tailwind code, or master prompts. Designed for designers, PMs, and founders, Mowgli delivers faster, more reliable design-to-development handoffs.
An AI-powered add-on for Google Sheets and Excel has been introduced to automate financial modeling and spreadsheet analysis directly within workbooks. ModelMonkey uses Anthropic's Claude Sonnet 4.5 to build 3-statement models, DCF valuations, and cap tables through natural-language prompts, while pulling data from SEC filings, web sources, and data warehouses. The platform autonomously executes multi-step workflows, writes dynamic and auditable formulas into cells, and creates charts while applying firm-specific modeling conventions. ModelMonkey explains every change and formula it generates, providing one-click undo functionality so analysts and finance teams can review and trust edits without copying and pasting between separate tools. The add-on also debugs existing formulas, ensuring accuracy and maintaining workflow continuity for financial professionals.
Photorealistic architectural visuals can now be generated in seconds from 3D models, sketches, or photos through a new AI-powered rendering platform. ArchRender uses cloud AI to interpret geometry, scale, and perspective, then applies materials and lighting while stylizing scenes based on natural-language prompts or uploaded moodboards. The platform enables designers and architects to bypass complex lighting and setup workflows, instead iterating rapidly with controls for seasons, times of day, edit modes, and design variants. ArchRender exports high-resolution upscaled renders suitable for client presentations, transforming the traditional rendering process that previously required extensive technical setup and rendering time.
A new AI platform converts home photos into photorealistic interior and exterior redesigns within seconds. HouseGPTs applies generative models to propose style variations, 3D renderings, virtual staging, and furniture and color suggestions from a growing design library. Homeowners, designers, real estate professionals, and flippers visualize renovation ideas and stage properties instantly, cutting weeks of wait time and architect costs while accessing realistic, editable concepts.
