Ever look at a robot and think, "I could eat that"? Swiss engineers did, apparently, and now we have a fully edible soft robot powered by ingredients from your kitchen pantry. Meanwhile, a smart bandage is diagnosing wounds and prescribing treatment without human intervention, glowing bacteria in a pill could make colonoscopies optional, an AI lab assistant is running more reactions per month than grad students do in years, and a menstrual cup finally got its PhD in data science. Welcome to the future, where everything is either edible, AI-powered, or both.
Table of Contents
🤖 A robot you can actually eat (and it tastes like gummy candy)
NEWS
Imagine handing a wild boar a robot that wiggles like prey, and when the boar eats it, it gets vaccinated. That's not a fever dream from a particularly creative grant application. It's a real thing that researchers at EPFL just built.
The team created what appears to be the first 100% edible soft robot with an integrated power source. The pneumatic battery uses citric acid (the sour stuff in lemons) and sodium bicarbonate (baking soda), separated by a membrane. Apply pressure, the membrane breaks, carbon dioxide is generated, and the robot bends rhythmically for several minutes. The actuator itself is made from gelatin and glycerol, essentially a functional gummy bear that can move on its own.
Lead author Bokeon Kwak describes the taste as "a little bit sweet" for the actuator and "crunchy on the outside and sour on the inside" for the battery. The team has already served flavored versions to humans at Expo 2025 Osaka as part of their RoboCake demonstration, featuring edible robotic teddy bears alongside chocolate batteries powering LED birthday candles.
One of the practical applications here is wildlife management. Traditional vaccination requires capturing animals or distributing bait by air. An edible robot that mimics live prey could lure skittish animals like wild boars, deliver vaccines or nutrients, and leave zero toxic waste behind. The EU-funded RoboFood project backing this work has €3.5 million to prove the concept.
Director Dario Floreano sees this as foundational technology for all biodegradable robots, because, of course, the replacement for lithium batteries might be baking soda.
🩹 This smart bandage diagnoses wounds and prescribes its own drugs
RESEARCH
Your future doctor might be a bandage. Engineers at UC Santa Cruz built a-Heal, a closed-loop wound healing system that photographs injuries, diagnoses their stage using machine learning, and automatically delivers the optimal treatment without human intervention.
The DARPA-funded device combines a miniature fluorescence camera, AI algorithms, and bioelectronic actuators that can deliver both electric fields and drugs directly to the wound. Every two hours, it captures 11 images at different focal depths, runs them through an algorithm called Deep Mapper to determine which healing stage the wound is in, then decides whether to apply electrical stimulation or release fluoxetine (yes, the antidepressant) to accelerate recovery.
In pig models (chosen because pig skin closely resembles human skin), treated wounds showed 25% faster healing compared to standard care. Re-epithelialization hit 51.8% on treated wounds versus 15% on controls. The system detected inflammation markers and automatically switched treatment modalities at the appropriate time.
The technical achievement is nothing to be scoffed at. A reinforcement learning agent that learns from each individual wound rather than applying generic protocols. Professor Marco Rolandi describes it as "an automated personalized physician" that optimizes treatment as healing progresses.
Chronic wounds affect 6.5 million Americans annually, incurring costs up to $25 billion. A bandage that monitors continuously and adjusts treatment in real-time could transform outcomes for diabetic ulcers and other stubborn injuries. Turns out the bandage was always the smartest person in the room.
💊 Glowing bacteria in a pill might make colonoscopies optional
RESEARCH
Nobody likes colonoscopies. Researchers in China apparently took that personally and developed an ingestible bacterial sensor that glows when it detects blood in your gut, potentially eliminating the need for invasive procedures in some patients.
The MagGel-BS platform encapsulates genetically engineered E. coli Nissle 1917 (a well-established probiotic strain) in sodium alginate hydrogel microspheres containing magnetic particles. Swallow the pill, wait for it to pass through your system, retrieve the microspheres from your stool using a magnet, and measure how much they glow. The light intensity correlates with disease severity.
The bacteria are programmed to emit light when they encounter heme, the iron-containing compound in blood. Dr. Ying Zhou and colleagues at East China University of Science and Technology achieved detection within 20 minutes (total analysis time around 25 minutes), compared to hours for traditional biosensors. The hydrogel shell solved two longstanding problems: it protects bacteria from digestive acids and makes retrieval straightforward.
Important caveat: this is diagnostic only. The system detects gastrointestinal bleeding associated with IBD, colorectal cancer, and similar conditions, but cannot visualize tissue or remove polyps. In mouse studies, the microspheres showed no immune responses or adverse effects and proved biocompatible.
Human trials haven't started yet, but Dr. Alan Moss from the Crohn's & Colitis Foundation called the concept "interesting" and noted that fecal calprotectin and colonoscopy both have limitations for monitoring IBD patients. A magnetic poop retrieval system might not sound glamorous, but neither does bowel prep.
🧪 This AI chemist runs more experiments than PhD students do in years
RESEARCH & NEWS
Drug discovery has a bottleneck, and it's not biology. It's making the molecules. OnePot AI emerged from stealth with $13 million to fix this, using an AI agent named Phil that can design, execute, and analyze chemistry experiments autonomously.
Co-founders Daniil Boiko (CMU PhD candidate, co-creator of the Coscientist AI system published in Nature) and Andrei Tyrin (MIT computer science graduate) built a custom synthesis lab called POT-1, where Phil directly controls liquid handlers, analyzes mass spectrometry data, identifies unexpected byproducts, and generates new hypotheses to test. “In one month, Phil ran more reactions than a typical graduate student completes during their entire PhD” - from their own introduction.
The investment came from Fifty Years, Khosla Ventures, and angel investors, including OpenAI co-founder Wojciech Zaremba and Google Chief Scientist Jeff Dean. The company offers a library of 1.9 million possible compounds, many never synthesized before, across five core reaction types, including Suzuki-Miyaura coupling and Buchwald-Hartwig amination.
Traditional contract synthesis takes 6-12 weeks for a handful of molecules. OnePot claims 6-10x faster delivery by capturing every experimental detail for perfect reproducibility. Their internal benchmarks suggest Phil's predictive models rival or exceed seasoned organic chemists.
The pitch is blunt: "AI's next frontier is not chatbots (it is atoms)." With 10^60 drug-like molecules theoretically possible and only 122 million compounds existing in PubChem, Onepot the synthesis problem isn't going away. But an AI that runs experiments at 2 AM while generating hypotheses might finally close the gap between computational drug design and physical reality. the synthesis problem isn't going away. But an AI that runs experiments at 2 AM while generating hypotheses might finally close the gap between computational drug design and physical reality.
🩸 Smart menstrual cup tracks what period apps can only guess
NEWS
Five years, thousands of design iterations, and now Bristol-based Emm has raised $9 million to launch what it calls the world's first working smart menstrual cup. Previous attempts (notably Looncup's 2015 Kickstarter) promised similar features but never shipped functional products.
Founder Jenny Button started the company during the COVID lockdown after noticing she wore multiple health wearables but had no device tracking reproductive health. She recruited former Dyson engineer Chris van Kempen to solve the engineering challenge: ultra-thin capacitive biosensors embedded in medical-grade silicone that measure fill level, flow volume, and cycle patterns without transmitting data while inside the body.
The cup tracks daily flow rate, total menstrual fluid loss per cycle, period length, and cycle regularity. Data transfers only when the device is outside the body and synced with its UV-sterilizing charging case, Emm then connects via Bluetooth to an app that builds baseline profiles over three cycles. Lunar Ventures led the round, with participation from Labcorp Venture Fund and angels including former Oura CEO Harpreet Rai.
Menstruation is technically the fifth vital sign, but endometriosis diagnosis still takes 7-10 years on average due to poor data characterization. Emm's explicit goal is to accelerate diagnosis by giving clinicians objective measurements rather than patient recall.
The UK launch is scheduled for early 2026, with US availability expected around 2027. Over 30,000 people on the waitlist suggest demand exists for menstrual health data that doesn't require manually logging symptoms into an app. Privacy architecture ensures that even in a Roe v. Wade scenario, Emm couldn't hand over user data because they don't store it in an identifiable form.
We live in remarkable times when the most sophisticated AI systems are being deployed to heal wounds, detect gut disease, synthesize drugs, track periods, and feed medicine to wild boars via edible gummy robots. The common thread across all five stories is measurement: better data, captured automatically, analyzed intelligently, leading to actions humans either couldn't take or wouldn't think to take.
If any of these developments made you reconsider what wearables, pills, or bandages could do, that's precisely the point. Science keeps expanding the definition of "medical device" while shrinking the gap between diagnosis and treatment. The real question isn't whether we'll accept robots we can eat or bacteria that glow in our intestines (it is how quickly).
Share this with someone who needs to know that baking soda robots are real and functional. Forward it to the friend who thinks femtech is just period tracking apps. Send it to the chemist who's been manually running reactions for months.