In Sunday's digest of oddities, somewhere between a robot that can't be ticketed and a court that cited the wrong statute, a brief news item nestled in: scientists have baked cookies from plastic bottles. With vanilla grown by yeast. Called µBites — "microbites," a play on words: micro-particle and "bite." It was shown on August 24 at the American Chemical Society meeting in Chicago, and any normal person's first reaction is to grimace and scroll past.
I paused.
Because the news had two hooks at once. First: behind the cookies stands NASA — the Deep Space Food Challenge program, meaning this isn't a startup with a catchy name or an article in a dubious journal, but a task for which the agency handed out grants to three hundred teams.
Second: skeptic Jason Hallett from Imperial College said something you can't just pass by: "We produce 400 million tonnes of plastic waste a year. You're not going to turn all of that into cookies." True. But then why this cookie at all? That's the question I went looking to answer.
The technological chain for µBites is surprisingly short, with only two non-human performers: pressure and yeast. First, used PET bottles — physically real ones, bought at a supermarket — along with corn stalks and leaves are ground into a uniform slurry. Then it's poured into a reactor with a name from a geochemistry textbook: oxidative hydrothermal dissolution, OHD. Inside — water, oxygen, high temperature and pressure; output — a carbon-rich liquid in which nothing recognizable remains of the bottle. According to SIU's patent application, the process dissolves any carbon-containing waste with over 90 percent efficiency.
OHD wasn't invented by a food technologist, but by geology professor Ken Anderson from the same university — ten years ago, for completely different tasks converting biomass into fuel. This, incidentally, is the typical fate of space technologies: almost no one in this story originally set out to feed astronauts. The apparatus simply knew how to do one thing — turn dead carbon into liquid carbon accessible to living organisms. Then the yeast enters.
Yeast is humanity's oldest biotech factory: beer, bread, wine. Insulin, which used to be squeezed from the pancreatic glands of slaughtered cattle, is now brewed by them — programmable strains of S. cerevisiae. The team of microbiologist Lahiru Jayakody and graduate student Sandhya Jayasekara did with yeast what pharmaceuticals did forty years ago: rewrote their profession.
Three strains participate in the work. Baker's yeast and probiotic S. boulardii drive proteins, fats and acids from the carbon liquid. A separate engineered S. cerevisiae produces vanillin from ferulic acid in plant biomass — a flavoring so the cookie smells like food, not a laboratory. The boldest is the third strain, Rhodosporidium toruloides: through adaptive laboratory evolution, meaning months of selection on a meager diet, it was taught to eat ethylene glycol — the toxic component into which PET breaks down — and build β-carotene from it, that same carrot pigment from which our bodies make vitamin A.
So the chemistry here isn't far-fetched. Vanillin from plastic was first obtained back in 2021 in Edinburgh: Joanna Sadler and Stephen Wallace fed terephthalic acid to modified E. coli, and within 24 hours it converted 79 percent of the raw material into vanillin — because the molecules are chemical near-neighbors, the microbe just needs to rearrange a few hydrogen and oxygen atoms on the ring. The SIU team took the proven idea, transferred it to yeast, and brought it to testing in an oven. The resulting dough with added starch and sweetener goes into a 3D printer, which extrudes cookie discs.
To understand why all this was undertaken, you need to rewind to January 2021. NASA and the Canadian Space Agency launched the Deep Space Food Challenge — a competition for food systems for deep space. The condition sounds simple: the system must work three years without replenishment and feed a crew of four. Not necessarily the entire diet — diverse nutritious products. Over three years, more than 300 teams from 32 countries passed through the competition.
The finale took place in August 2024 at Ohio State University, where four American teams ran their systems under combat conditions for two months. Four students managed the tests — NASA officially called them "Simonauts" — and the jury evaluated safety, sensory qualities, edibility and production volumes. The $750,000 grand prize went to Interstellar Lab from Florida with the NUCLEUS system: autonomous phytotrons, vegetables, microgreens and insects for micronutrients. The international winner was Finland's Solar Foods with gas fermentation. And the µBites team with a first-phase grant of $25,000 was no longer among the finalists.
And here's an interesting twist that press releases try not to emphasize: SIU dropped out of the competition at an early stage, but the project didn't die. In 2024, Jayakody received an NSF CAREER grant for $624,500 — twenty-five times more than NASA gave — and brought the technology to the ACS presentation. So formally µBites is an "out of competition" project; from NASA they got only the task conditions and a starter check.
The question "why feed people trash" disappears if you pick up NASA's calculation of a transit food system. Mars mission: 9,660 kilograms of food for a crew of six for 1,000 days — almost ten tons, of which 1,440 kg goes to packaging alone. And this is cargo that must not just be lifted from Earth, but preserved: vitamin C in space rations degrades to zero after a year of storage, vitamin A and folic acid melt away next. The American food system on the ISS is designed for one and a half to three years; for a Mars mission you need a five-year one, and according to old NASA estimates, almost no items survive to the fifth year.
Against this backdrop, the logic of µBites reads without poetry: every kilogram of food the crew makes on site from their own trash bin is a kilogram that doesn't need to be transported, packaged and heated. The system's service life in the patent is written with the same arithmetic — minimal maintenance for three years. That's exactly how long a Martian "round trip" takes, according to Jayakody. Earth applications — submarines, polar stations, disaster zones — are already pleasant appendages to the same machine.
The price deserves a separate honest word. According to The Guardian, a kilogram of µBites currently costs $60. That's hundreds of times more expensive than flour and tens of times more than good chocolate. No one on the team hides this: economics will only appear along with scale and yeast efficiency.
For now, the product isn't edible even for its creators — there have been no tastings yet, the team is waiting for institutional approval for human trials. Survey participants only rated the cookie by smell, and it smells good, they say; most said "would eat in resource-scarce conditions." One BBC journalist put it most honestly: "if I were in a disaster zone and had nothing to eat — I'd eat it. But from the shelf instead of Oreos I wouldn't choose it."
The idea of making protein from non-food was by no means born at NASA. The largest-scale experiment in history was conducted by the USSR: paprin, a protein-vitamin concentrate from yeast grown on paraffins — petroleum refining waste. The first plant in Kstovo (1973) produced 35 thousand tons per year, and by 1987 the entire industry produced about 1.1 million tons of PVC — two-thirds of global volume. A feed additive with 56-percent protein from oil — sounds like a quote from dystopia, but animal husbandry experts accepted it.
The story ended characteristically: after an accident with emissions at the Kirishi Biochemical Plant in 1987, an environmental movement swept across the country — from Kirishi to Angarsk — and by the mid-nineties production died. The paprin lesson isn't "the idea is bad," but "scale without a risk map is fatal." Finland took both circumstances into account in its version: Solar Foods grows Solein protein from CO2, water and electricity, the first commercial factory in Vantaa produces up to 160 tons per year — quite modest, but with a full regulatory package: approval in Singapore since 2022, FDA application filed in 2025. The Factory 01 bioreactor produces as much protein per day as dairy protein from a 300-cow farm — only without cows, fields or weather.
There's also a third, most poetic route: New York's Air Company at the same competition built a system that takes carbon dioxide from the astronauts' own breath, converts it to alcohol, and feeds yeast with the alcohol. The result is a protein shake that, according to the co-founder, tastes "as good as seitan." Food from crew breath — a closed cycle that science fiction writers wrote about. Space has always been Earth's most expensive marketing channel: Tang, which everyone remembers as "NASA's drink," was actually developed by General Foods back in 1957 and sold poorly until John Glenn drank it in orbit in 1962. µBites, if they make it to the supermarket, will follow the same track: first "astronaut food," then just a product.
Now back to Hallett from Imperial College. His arithmetic is impeccable: according to OECD data, the world produces 460 million tons of plastic per year and discards 353 million tons, of which 9 percent is recycled. Even if every µBites cookie consumed a bottle, against the backdrop of this flow the technology is a drop in the ocean, and it will never become a "solution to the plastic crisis." Hallett formulates it harsher: "commercially it won't take off. We're not going to eat plastic cookies."
But here it's important to notice the substitution made by both the skeptic and part of the press. µBites was never positioned as a solution to the waste crisis — that's a side effect that the team likes to mention in grant applications. The main task is formulated directly in the presentation abstracts: "protein-rich supplements." And the scale of the target application isn't 8 billion people, but 9.7 tons of food for six people for three years. For such an audience, the productivity of a yeast reactor isn't laughable, but sufficient. Compare: the full cycle of processing global plastic requires trillions of dollars, while the full food cycle of one Martian expedition requires one reactor, a rack of bottles, and patience.
An honest rough edge that remained with me after the entire breakdown: not one source — not in the ACS press release, not in The Guardian, not in the abstracts — says what proportion of the cookie's mass actually comes from plastic. Proteins and vanilla from waste — yes; but the starch and sweetener binding the dough are still ordinary additions, and the team only "hopes" to microbize them in the future. A cookie from plastic is still a cookie with plastic, and the difference between these formulations is roughly like between "a house of brick" and "a house with brick."
The second rough edge: long-term safety of such products doesn't exist for anyone. Microplastics are already found in human brains, blood and placentas — we, whether we like it or not, eat plastic bit by bit every week, for free and without vanilla. µBites at least break it down to molecules. But safety will have to be confirmed by clinical trials, and their turn hasn't come yet.