Psst… We received a tip…

You bag up the clothes you stopped wearing and feel briefly virtuous. The faded band tee, the gym shorts that pilled, the fleece that lost its zip. Into the donation bin they go, off to a second life. That is the story we tell ourselves at the bin.

Almost everything you wore was a blend. The tee was cotton spun with polyester, dyed, then finished with something to stop it wrinkling. The shorts were polyester laced with elastane so they stretched. To a recycling line, that’s a black box. Machines can shred it. Pulling clean materials back out so they become clothing again is the hard part, and mostly it stalls right there.

A recent patent out of Vancouver wants to change what happens after the bin. Two red shirts that look identical can need two different recipes, and the system is built to notice the difference.

The patent was authored by Sixone Labs, a recycling startup based in British Columbia. It targets textiles that are mixed, blended, or chemically messy.

Sixone's system tackles this.

First, a sensor (the patent leans on infrared spectroscopy) captures a spectrum of the fabric, a kind of chemical fingerprint. Then a machine-learning model compares that fingerprint against everything it has seen before and predicts a processing protocol. Which reaction, what temperature, which solvent, at what concentration, for how long? The fabric gets treated under those conditions, usually by depolymerisation, which breaks polyester back into the small molecules it was built from. Those monomers can be repolymerised into fresh pellets.

Then the useful trick. The system measures how well the reaction actually went, things like yield and purity, and feeds that result back into the model as new training data. The patent describes the knowledge base as fluid rather than fixed, so the categories it sorts materials into shift as it learns. 

Get the loop turning on enough garments and the claim is that you can predict a working recipe for a fabric the machine has never seen, without a full lab workup each time.

Textiles are one of the worst-performing waste streams on the planet.

The world generates roughly 92 million tonnes of textile waste a year, and under 1% of it gets recycled back into new clothing, per researchers publishing in Science Advances. Most is landfilled or burned.

The reason is chemistry rather than laziness. Mechanical recycling (the cheap and common method) mostly needs clean single-fibre inputs like pure cotton or pure wool. Blended fabrics defeat it. And the "recycled polyester" already in your wardrobe is misleading. Around 98% of it comes from plastic bottles according to Textile Exchange. The industry has been recycling drink containers and calling it circular fashion.

So the real bottleneck is blends. Until someone can take a poly-cotton-elastane shirt and pull real, reusable polyester back out at a sensible cost, the donation bin is mostly a slower route to landfill.

Textile-to-textile recycling has gotten crowded, and it splits into two camps that Sixone's patent straddles.

One camp does the chemistry. 

Circ, based in Virginia, has built a patented process that recycles poly-cotton blends and recovers both fibres, with Inditex (Zara's owner) and Patagonia among its backers. 

Ambercycle in Los Angeles and France's Carbios break polyester down to monomers, Carbios using engineered enzymes rather than solvents. Australia's Samsara Eco uses enzymes too and has a supply deal with Lululemon. Sweden's Syre, spun out with H&M backing, is going straight for industrial scale. 

These companies own real reactors and real offtake deals. What most of them run, though, is one optimised process applied to feedstock they have to keep clean.

The other camp does the seeing. 

Refiberd (hyperspectral cameras plus AI) and the UK's Matoha (handheld infrared scanners) identify what a fabric is made of in about a second, so it lands in the right bin. They sort. The transforming is someone else's job.

Sixone's patent sits in the gap as an AI that reads the fabric and then prescribes the chemistry, with a feedback loop tuning the recipe over time.

Textile-to-textile recycling was worth about US$2.8 billion in 2024 and is projected by Global Market Insights to reach US$44.8 billion by 2034, a 24.9% compound growth rate that says more about optimism than certainty. Widen the lens and the estimates scatter. The broader textile recycling market sits closer to US$8.4 billion in 2025, heading to US$11.9 billion by 2030.

The pressure behind the numbers is concrete. Polyester is now about 59% of global fibre production, and 88% of that is fossil-based, so most of what the world wears is plastic with no good end-of-life path. Regulation is closing that gap by force. In September 2025 the European Parliament finalised mandatory Extended Producer Responsibility for textiles, which makes brands pay for what happens to clothing after the sale. Once disposal becomes the brand's bill, a technology that turns waste into reusable polyester can really cut costs.

Money is moving into textile recycling faster than the technology is maturing.

The flagship bets are large and brand-driven. Sweden's Syre closed a US$100 million Series A in 2024, sitting on top of a US$600 million, seven-year commitment from H&M to buy its recycled polyester. Australia's Samsara Eco raised US$100 million to scale its enzymatic recycling and build facilities in Southeast Asia.

The mid-stage players are grinding. Virginia's Circ has pulled in around US$100 million in total, including an oversubscribed US$25 million round in March 2025 led by Taranis. Ambercycle has raised over US$55 million since 2015. On the equipment side, recycling-machinery maker Erema took a stake in Brisbane's BlockTexx in October 2025.

The AI-sorting layer, the camp closest to Sixone's sensing claim, is funded far more thinly. Refiberd raised a US$3.4 million seed and Matoha closed about US$1.9 million. The gap between those tickets and the nine-figure chemistry rounds shows where investors think the value, and the risk, concentrates: in reactors and offtake, well ahead of the software that decides the recipe.

The chemistry is the easy part to trust. The harder problem is that a private model makes chemical decisions at scale while nobody outside the company can check its work.

When the model is right, you get clean monomers. When it misclassifies, you get a botched reaction with wasted feedstock, degraded or contaminated output, and a batch of "recycled" polyester whose real quality rests on a guess made inside a black box. A model with shifting “recipes” is harder to audit from the outside.

Now scale it. A brand buys recycled polyester and stamps a sustainability claim on a hangtag. A regulator, under the new EU rules, wants proof the material is what it says. The actual basis for that claim is a proprietary classifier's confidence score on a spectrum reading.

For the person at the donation bin, nothing visible changes yet. For a brand staring down EU producer-responsibility bills, a system that prices and processes mixed waste per piece could be the difference between a recycling line that loses money and one that breaks even. 

This week's patent is US 12,640,238, titled "Methods and systems for treating a heterogeneous mixture of materials," authored by Sixone Labs Ltd.

Read the filing if you want to see how the feedback loop is actually claimed, reply and tell us whether you would hand your wardrobe to an algorithm, and come find us on Instagram and LinkedIn.