The grind is the product: why the thing you'd sell stays in the drawer

TL;DR

93% of Americans bought something secondhand in the past year, but only 54% sold anything, according to the seventh annual US recommerce survey published in 2025. Meanwhile retail analysts GlobalData estimate US households sit on $580 billion of products they no longer use. The gap between those two numbers was never about price. It is attention: the listing, the lowballers, the 40 messages, the no-show. That grind is precisely the work an AI agent deletes, which is why agent-native commerce lands on the sell side first.

There is a graphics card in a drawer behind me. I replaced it, I know roughly what it is worth, and I have thought "I should sell that" maybe a dozen times. It is still in the drawer. If you run a homelab, some version of that sentence is true for you too, and I want to argue that neither of us is lazy. We are doing the math correctly, and the math is the problem.

Everyone buys used. Barely half of us sell

Buying secondhand stopped being a niche a while ago. The 2025 recommerce survey, the industry's own seventh annual read on the market, found that 93% of Americans bought a secondhand item in the past year, and projects the US recommerce market at $306.5 billion by 2030, about 8% of total retail.

The same survey found that only 54% sold anything.

Sit with that gap for a second, because it is the whole argument. Nearly everyone participates in this market as a buyer. Barely half participate as sellers, and every buyer needs a seller. Two decades of software made buying used effortless: search, filters, saved alerts, one-tap checkout. The direct sell side still runs on the same manual pipeline it always did. The friction of trading directly with another person is not spread evenly across the trade. It is concentrated, almost entirely, on the person holding the thing.

The drawer is not laziness, it is arithmetic

GlobalData's research with the resale industry put a number on what that concentration strands: $580 billion of products that American households own and no longer use. That is $4,517 per household, spread across 184 items. For electronics alone, a 2024 SellCell survey of more than 2,000 people estimated 1.3 billion unused devices in US homes, worth approximately $67 billion, with 88.5% of Americans keeping outdated devices they no longer use.

The national number looks irrational. The household decision behind it is not. Selling one item means an evening of photographing and writing, then days of intermittent answering, then coordinating with a stranger who may not appear. Against that, the expected return on a single mid-value item is small. For any one object, expected price minus attention cost comes out negative, so the rational move is the drawer. The $580 billion is just that correct little decision multiplied by 130 million households.

Trading directly was supposed to be the efficient market. The one who no longer wants the thing, the one who does, no middleman taking a cut. The economics of the trade are fine. The economics of the transaction are terrible, and the transaction is priced in a currency you cannot discount: your attention.

Why is selling used hardware such a hassle?

You already know, because you have lived it. The listing you rewrote twice because the first draft undersold it. The price you researched and were immediately told was wrong by someone whose opening offer was half of it. The availability question, asked and answered, asked and answered. The buyer who confirmed twice and did not come. Somewhere at the end of it, a car park.

The research measures the same thing from the outside. When the SellCell survey asked people why their devices stay unsold, the answers were not about money:

Valuation doubt, data anxiety, and not knowing where to start. Each one is an attention cost. None of them is a price problem.

Researchers at the Rochester Institute of Technology and the University of Waterloo surveyed 4,000 Americans about what happened to their last unused device and found that 39% simply stored it, while only about 1 in 10 resold it. Their sharpest finding is about timing: the worries that block the handoff barely register in the abstract and kick in "only when someone is facing the real decision". Intentions survive right up to the moment the process starts, which is exactly where the grind lives.

The price was never the problem

If price were the barrier, known prices would move items. They do not. Material Focus, a UK non-profit that surveyed 3,001 householders, counted 39 million unused tech items hoarded in UK homes, including working laptops worth £1.5 billion in total resale value.

A working laptop resells for £274 to £420. That is not a mystery price, and it is not a trivial one. The item sits anyway, which tells you the cost blocking the trade is not monetary. Everything in the drawer has a price. None of it has your attention.

The category's answer: automate the answering machine

The people who run today's marketplaces can read the same surveys, and in 2026 the largest listing platforms started shipping AI auto-replies for exactly that availability message. Toggle it on, and an assistant answers "is this still available" on your behalf, from the listing text.

I do not think this is a foolish move. From inside an architecture built for a person scrolling with a thumb, absorbing message number 40 is the available improvement. But notice what remains yours: writing the listing, setting the price, reading the lowball, judging the counterparty, holding your floor through the back and forth, and standing in the car park. The chat box automates the one step that was already the cheapest, because the system's unit of work is still a human reading a screen. A chat box bolted onto human-native pipes is not an agent. It is a better doorbell on a house you still have to run.

What an agent actually deletes

The best negotiator has no ego, infinite patience, and perfect memory of what things actually sell for. That has never been a person. It is, rather precisely, a description of software.

Hand the sell side to an agent and the grind does not get easier. It stops being yours. You state what you are selling and your floor. It handles the description, fields every availability question without resentment, reads an opening offer at half ask as information rather than insult, checks the counterparty's history before engaging, and walks away from a pressure tactic at 2am because a deadline invented to rush you is not a deal. You approve the outcome that matters. Nothing closes without you.

And on the other side of the table sits another agent doing the same job for its owner, which is what keeps this from being predation. I wrote separately about why symmetry is the whole category; the short version is that when both sides are tireless and both hold the comps, the price lands near what the thing is worth, and neither person spent an evening getting there. This is one instance of the broader shift toward software transacting over structured data, where agents buy what they can read.

That is agent2agent commerce, and the reason it starts sell side is the whole argument above: the seller is the one paying the attention cost today, so the seller is the one an agent unburdens first.

Both doors: the hunt and the drawer

Klodi is being built as exactly that: an Agent2Agent marketplace where your agent represents you and the counterparty's agent represents them. I will not pretend the hard parts are done, and I am not asking you to trade today. I am asking you to check the argument, and the code is public precisely so the argument can be checked.

So, two questions, and if you run an agent host they have the same answer. What are you hunting, and what is sitting in your drawer?

FAQ