Industries / Energy, water and waste / Environmental services / Recycling
Metals and glass
This work melts scrap metal and used glass back into new metal and glass, and recovers metals such as zinc from factory dust.
5
Milestones achieved, from 1977 to 2025.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1977
Bottle banks open
On 24 August 1977 the UK's first bottle banks, large street bins for used glass bottles and jars, opened in South Yorkshire and Oxford, with Barnsley starting that morning. The Glass Manufacturers' Federation ran the scheme, and its president, Stanley Race, chairman of Redfearn Glass, dropped in the first jar. About 500 metric tons were collected across the country in six months; by 1996, 380 local authorities took part, and by 2002 the UK had about 50,000 banks. In 1979 the government told Parliament that it encouraged the scheme but left it to councils.
Credit
Stanley Race and the Glass Manufacturers' Federation, now British Glass
1980s
Furnace dust becomes zinc
Electric arc furnaces melt scrap into new steel, and their fumes leave a dust rich in zinc that is handled as hazardous waste. Horsehead says its predecessor company built the market for recycling this dust from mini-mills, scrap-based steel mills, in the early 1980s, and that the US EPA named its process the best demonstrated available technology for the dust. Its Waelz kilns, rotating tubes about 49 m (160 ft) long heated above 1,100 °C, drive the zinc off as an oxide that zinc smelters use; Horsehead processed 579,000 tons of dust and other zinc-bearing material in 2015.
LessonBefesa, which now runs kilns at the same four US sites, reports treating 1,215,000 metric tons of steel dust worldwide in 2025 at an average 70% of capacity.
Credit
Horsehead Corporation and its predecessors
2016
A zinc refinery stalls
Horsehead spent about US$575 million on a plant at Mooresboro, North Carolina, to turn the zinc oxide from its dust kilns into zinc metal by chemical leaching and electrolysis, replacing a smelter at Monaca, Pennsylvania, that it shut in April 2014. The plant started in May 2014 but ran at about 25% of its capacity of 155,000 tons a year in late 2015; Horsehead idled it on 22 January 2016 and filed for bankruptcy on 2 February 2016.
LessonHorsehead reported an undersized treatment section that held up production, corroded electrodes, and pumps and lines designed wrongly for the job; Befesa bought a US zinc refinery in September 2022 and lists Mooresboro among its refining sites.
Credit
Horsehead Holding Corp
2025
Copper's recycled share holds steady
Copper can be melted and refined again without losing quality. The International Copper Study Group (ICSG), an intergovernmental body, estimates that recycled copper, counting clean scrap melted directly by fabricators and scrap refined at smelters, supplied about 11.8 million metric tons in 2025, 34.0% of all copper used, against 7.3 million tons and 33.8% in 2005.
LessonRecycled tonnage grew by about 60% in twenty years while its share stayed near one third, because demand grew too and much copper stays in buildings and power networks for decades.
Credit
ICSG, Copper Recycling Survey
2025
Oregon loses its bottle plant
O-I Glass told Oregon regulators on 30 July 2025 that it would permanently close its Portland bottle plant, open since 1956 and the only one in the state, as part of a cost-cutting plan. The plant had used more than 80% recycled glass in 2021, and the Glass Packaging Institute wrote in 2022 that O-I bought nearly 100,000 tons a year of glass collected in Oregon, about half of it from curbside programs.
LessonWhen the one nearby furnace closes, a region's collected glass loses its main buyer.
Credit
O-I Glass; reported by Resource Recycling
Watch for
Mooresboro makes zinc again
Befesa says it filed a claim with the US Internal Revenue Service in October 2026 for a production tax credit, under Section 45X of the Inflation Reduction Act, on critical minerals produced at its Mooresboro zinc refinery, the plant Horsehead idled in 2016. It estimates the credit at about US$29 million for 2026.
Achieved whenBefesa publishes the plant's yearly zinc output and the IRS accepts the claim, or another outside party confirms steady production.
Claimed by
Befesa, October 2026
Ohio glass back to bottles
Cullet LLC bought Repeat Glass, which collects glass from businesses around Cleveland and Akron, and says its new Ohio unit will collect glass separately and process it into manufacturing-grade cullet (crushed glass ready to melt) for new containers.
Achieved whena container glass plant confirms that it buys and melts the Ohio cullet, with tonnages.
Claimed by
Cullet LLC, March 2026
Open
01
Where arc-furnace dust goes
For every US electric arc furnace steel mill, a table of how much furnace dust it shipped in each reporting year and where it went: a zinc recycler's kiln, a treatment-and-landfill site, or somewhere else. Build it from EPA's public hazardous waste reports, and publish the share recycled by mill and by year.
The briefCredit · unclaimed02
Furnace-ready glass by route
A public, repeatable measurement of how much glass from each way of collecting it becomes furnace-ready cullet, meaning clean, sized and sorted well enough to melt into new bottles: deposit returns, glass collected separately at the curb, drop-off bins, and glass out of single-stream sorting plants. Weighed samples are hand-sorted and sieved, and the method, the yields with error bars, and photos of every sample are published.
The briefCredit · unclaimed03
Measured copper in scrap wire
An open table of the measured copper yield, copper weight divided by whole cable weight, for the common kinds of scrap insulated wire, from stripping and weighing hundreds of samples, plus a phone tool that reads a cable's printed jacket and photo and estimates its yield.
The briefCredit · unclaimed
Open 01 · the brief
Where arc-furnace dust goes
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
For every US electric arc furnace steel mill, a table of how much furnace dust it shipped in each reporting year and where it went: a zinc recycler's kiln, a treatment-and-landfill site, or somewhere else. Build it from EPA's public hazardous waste reports, and publish the share recycled by mill and by year.
Why it is still open
The recyclers publish only their own totals, such as Horsehead's 579,000 tons in 2015 and Befesa's 1,215,000 metric tons worldwide in 2025. EPA's Biennial Report records, for each large generator, the quantity of hazardous waste and whether it was sent for recycling, treatment, storage or disposal, and offers the national data as downloads, but a search found no mill-by-mill analysis of where the dust goes; a 2026 analysis of the 2021 and 2023 Biennial Report files on GitHub ranks waste codes and generators nationally and does not follow the dust to its destinations.
Built with what exists
EPA's Biennial Report downloads and its public e-Manifest records, filtered for K061, the federal waste code for "emission control dust/sludge from the primary production of steel in electric furnaces"; Global Energy Monitor's Global Iron and Steel Tracker for the list of mills; Python and DuckDB; a coding agent that writes and tests its own checks, such as one built on Inkling 2 Small (see the Note "Checked without being asked").
First step
Download the latest Biennial Report files, filter for K061, and count the generators that ship it and the sites that receive it.
Done looks like
Every US arc furnace mill in the tracker matched to its EPA ID number, with dust tonnage and destination for each reporting cycle from 2015 to 2023. National totals agree within 10% with the recyclers' own published throughput for the same years, and 20 mills are checked by hand against their permits or reports.
Open 02 · the brief
Furnace-ready glass by route
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, repeatable measurement of how much glass from each way of collecting it becomes furnace-ready cullet, meaning clean, sized and sorted well enough to melt into new bottles: deposit returns, glass collected separately at the curb, drop-off bins, and glass out of single-stream sorting plants. Weighed samples are hand-sorted and sieved, and the method, the yields with error bars, and photos of every sample are published.
Why it is still open
In 2024 the glass processor Strategic Materials said 98% of deposit glass, about 92% of glass collected separately at the curb and about 60% of drop-off glass was fit for new containers, without giving a method or a single-stream figure. The Container Recycling Institute's 2017 comparison of programs called its own findings "too preliminary to be considered true benchmarking" and relied on an industry estimate of about 65% usable cullet from single-stream plants. Oregon's environment department published a study in May 2026 of where each material entering the state's commingled sorting plants ends up, glass included, but almost all Oregon programs collect glass separately, so it does not compare routes. CulletIQ, a prototype posted on GitHub in October 2026, estimates the color mix from cullet photos with thresholds it calls uncalibrated, checked on synthetic images and two hand-sorted photos.
resource-recycling.comresource-recycling.comoregon.govgithub.com
Built with what exists
A set of test sieves, a bench scale, cut-resistant gloves and eye protection, and a phone camera with a color card; Ultralytics YOLO or the vision model in Liquid AI's LFM3 family, fine-tuned on one graphics card (see the Note "One GPU to fine-tune"), to estimate the mix from photos; a glass processor or sorting plant willing to share samples. The image set of trouble pieces proposed in glass-and-ceramics.md can serve as extra training data.
First step
Write the sorting protocol, ask one glass processor for five samples from each route, and sort them by hand.
Done looks like
At least 30 samples of at least 20 kg from each of four routes in one region, each sorted into furnace-ready cullet by color, fines below a stated size, ceramics, stones and porcelain, and non-glass. Yields per route are published with 95% confidence intervals and an open license, together with how far the photo estimates fall from the hand sorts.
Open 03 · the brief
Measured copper in scrap wire
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table of the measured copper yield, copper weight divided by whole cable weight, for the common kinds of scrap insulated wire, from stripping and weighing hundreds of samples, plus a phone tool that reads a cable's printed jacket and photo and estimates its yield.
Why it is still open
Scrap yards and recyclers publish "typical" ranges without saying how they were measured, and they disagree. A UK recycler, Dunmow Group, lists telephone, broadband and computer cable at 20% to 35% copper; a New Jersey yard, Rockaway Recycling, prices Cat 5/6 data cable as "45-50% ICW"; H&C Metals puts Romex house wire at 62% to 64%. A search found no measured, open dataset.
Built with what exists
Hand wire strippers or a small benchtop stripper, a scale reading to 0.1 g, calipers and a phone camera; a vision model to read jacket printing such as "12/2 NM-B" and classify the cable, for example an LFM3 model fine-tuned on one graphics card (see the Note "One GPU to fine-tune"); Ultralytics YOLO. Use only disconnected scrap wire.
First step
Collect 1 m lengths of ten common cable types, strip and weigh three of each, and post the first table.
Done looks like
At least 400 samples of at least 15 cable types from at least three sources, each with photos, jacket text, length, whole weight and copper weight, released under an open license. The mean and spread for each type are published, and the phone tool's estimate falls within 5 percentage points of the measured yield on a held-out set of 100 samples.