Why selling something online is such a hassle: the market kept the listing and dropped the talking

TL;DR

Selling something online is such a hassle because the market kept the listing and dropped the talking. Listing is now a photo and a suggested price, and the listing gets found. The back and forth that sets the price is still yours, and it is most of the work: in a study of 25 million bargaining threads on one of the largest online marketplaces, almost 80% of listings never received an offer, the first offer opened at 58 to 66 cents on the dollar, and bargained sales settled at 73 cents. An agent that does the talking for you, and walks when it should, is what puts the second verb back.

Why is selling used hardware such a hassle?

You know the sequence. The listing took four minutes. You picked the photos, the app suggested a price, you pressed post. Then the messages. An offer at half the number, sent to you and to twenty other people. The same question, asked and answered, asked again. A buyer who confirmed twice and did not come. Somewhere at the end, a car park.

None of that happened before the listing existed. All of it happened after. That is the whole argument, and it is why marketplaces feel the way they feel: the part they solved is the part before the listing. The part they left is the part that sets the price.

The market was always two verbs

The Greek agora was the centre of a city's business, social and political life, per the encyclopedia entry. From that one square the language got two verbs: agorazo, "I shop", and agoreuo, "I speak in public". Trading and talking. One place, two jobs.

The talking was the part outsiders noticed. When a Spartan herald warned Cyrus off the Greek cities, Herodotus has the Persian king answer with contempt for the square itself:

He was describing a market. What he saw was people talking prices at each other. Aristotle later proposed splitting the two jobs into two squares, a "freemen's agora" from which "all trade should be excluded" and "a traders' agora, distinct and apart from the other", in Politics VII.12. It stayed a proposal. For most of history the place you sold a thing was the place you argued about what it was worth.

Which verb did the marketplace keep?

Ronald Coase named the two costs of using a market in 1937. The first "is that of discovering what the relevant prices are". The second is "negotiating and concluding a separate contract for each exchange transaction", he wrote in The Nature of the Firm. Finding, then talking.

Online marketplaces took the first cost to nearly zero. The economists who study them say so in their definition: Liran Einav, Chiara Farronato and Jonathan Levin, in the Annual Review of Economics, describe "the primary function of these markets as making it easy for buyers to find sellers" and the design problem as matching them "while keeping search frictions low". A marketplace is a search machine, and a very good one.

Look at what the largest listing platform shipped in March 2026, in its own words. Upload photos and the assistant creates "a draft listing, filling in details, and even suggesting a price based on similar items listed in your area". It already carries more than 3.5 million new listings a day in the US and Canada. And it added an auto reply for one message: when a buyer asks whether the item is still available, the assistant answers from the listing. Sellers, as TechCrunch put it, "often don't want to spend excessive amounts of time responding to initial or redundant messages."

The one message the platform automates is the availability question, which is not a negotiation. It is the doorbell. The listing verb is now a photo. The talking verb is still a person, at 11pm, reading an offer at half.

	Verb one: list and be found	Verb two: talk and set the price
Who does it	The platform, from a photo	You
How long it takes	Minutes	Days, in messages
What it costs	Nothing you notice	Your attention, every evening it runs
What the platform shipped in 2026	A draft listing and a suggested price	An auto reply to "is it still available"
What decides the price	Nothing yet	This

What the talking actually costs, measured

The largest measurement of online haggling that exists is a 2020 study in the Quarterly Journal of Economics by Matthew Backus, Thomas Blake, Brad Larsen and Steven Tadelis. They took 88 million listings on one of the largest online marketplaces, the ones where a buyer could make an offer instead of paying the asking price, and followed 25.5 million bargaining threads. More than half the listings were used goods.

The listing verb worked: every item was posted and found. Then: "almost 80% of listings never receive an offer."

Being found is not being talked to. Four listings in five sat there.

Now the fifth, where the talking starts. The average asking price was $95. The first buyer offer opened at 57.5% to 66% of asking, depending on category. That is the lowball, measured across millions of people: the opening move is about 60 cents on the dollar. Once bargaining happened, the average sale settled at 73% of asking. Across all sales, bargained or not, it was 83%.

The talking is where a quarter of the price goes. Nothing about the listing decided that. The messages did.

And the talking often fails. One third of threads ended in immediate agreement. "The majority of sequences play out differently, ending in disagreement or delayed agreement." The odds depended on the goods: 59.6% for media, 31.7% for electronics, the lowest in the study. For the things a homelab operator sells, fewer than one thread in three that got as far as an offer ended in a sale.

One more detail says what the platform does about verb two. Each side was limited to three offers, and each offer expired after 48 hours. The platform did not do the talking. It put a clock and a cap on it and handed it back.

Why do marketplaces fail the seller? Because the second verb is priced in attention

Put the two verbs together and the drawer makes sense. In the seventh annual US recommerce survey, 93% of adults bought something pre-owned in the past year and 54% sold something. GlobalData's research with the resale industry put $580 billion of no-longer-used goods in US homes. When researchers at the Rochester Institute of Technology and the University of Waterloo asked 4,000 Americans what happened to their last unused device, 39% had stored it and about 1 in 10 had resold it.

The industry knows. Its own 2026 report, fielded by GlobalData across 3,268 US adults, says the next $23.3 billion of US resale arrives when "the act of selling" is "just as easy as clicking 'buy'", per the report's release. One of its executives split the seller's friction in two for Preloved Pod: the mechanical part, photos, pricing, shipping, and "that internal friction of asking, 'is selling this actually worth my time?'". The mechanical part is verb one, nearly solved. The question is verb two.

Four listings in five get no offer. The fifth opens at 60 cents. The evening it costs is certain. The sale is not. For any one item, the return minus the attention comes out negative, so the thing stays in the drawer. That is not laziness. It is arithmetic, and the arithmetic was never about the price.

What an agent does with the second verb

Here is why this was always going to be hard. Verb one scales because a listing is data: a photo, a title, a price, a location. Verb two never scaled because it is a conversation between two people who each know something the other does not, and it takes a person's attention for as long as it runs. A platform can put a clock on it. It cannot have it for you.

An agent can. Not by answering the doorbell faster, but by having the conversation. Your agent opens the channel, makes and reads structured offers, and holds your number without ever stating it. It knows what the part has been closing at. It reads "final price, gone tonight" as a pressure tactic, because it is one, and declines. If the right deal is three weeks out, it waits three weeks. When the offer is wrong, it walks away, the move a person finds hardest after a week of messages.

Three things you get, said plainly. Your negotiating policy, your ceiling, your floor and where you walk, stays on your own disk; only the offers and the listings travel. Nothing closes without you: every deal ends at a human approval gate, so you make one decision instead of forty. And the other side has an agent too, so nobody is outgunned and nobody is grinding. That symmetry is what Agent2Agent commerce means.

That is what Klodi is being built as: an Agent2Agent marketplace, where your agent represents you and the counterparty's agent represents them. Six host adapters, one identity, Apache-2.0, and the plugin tree is public. The hard parts are not done, and this piece is not asking you to trade today. It is asking you to check the argument against the code.

Both doors

Two questions, and if you run an agent host they have the same answer. What are you hunting? What is sitting in your drawer? The person with a spare card in a drawer is the person hunting a second one for the next build, and one install serves both. Verb one was never your problem. Hand over verb two.

FAQ