---
title: every market starts empty
canonical_url: https://ensurance.app/guide/every-market-starts-empty
markdown_url: https://ensurance.app/guide/every-market-starts-empty.md
subtitle: "the cold start problem, and why the first buyer has to be us, in public"
category: ensurance
---

# every market starts empty

*the cold start problem, and why the first buyer has to be us, in public*

Anyone who has tried to build a marketplace meets the **cold start problem** before they learn its name. Buyers will not come until there are sellers. Sellers will not list until there are buyers. A ride app with no drivers loses its riders, and a ride app with no riders loses its drivers. Economists call this a **two-sided market**, and one of the papers that framed it put the trap in its title: *Chicken & Egg* (Caillaud and Jullien, 2003).

Andrew Chen, who led rider growth at Uber, wrote a whole book on it. *The Cold Start Problem* (2021) argues that a networked product is worth little to its first users, because the network that gives it value is not there yet. Every two-sided market has to get from zero to the first useful crowd on both sides, and nothing about the product does that on its own.

## the trap

### what is the cold start problem?

The cold start problem is the stretch at the start of any two-sided market when neither side has a reason to show up, because the other side is not there yet. It ends only when someone fills one side first.

## a forest with nine million dependents

Now put a forest on one side of that market.

The Catskill and Delaware watersheds in upstate New York are forests, streams, and reservoirs that supply drinking water to about nine million people in and around New York City. The forest filters that water whether or not anyone pays it to. When the city weighed building plants to do the same job, filtration plants for the Catskill and Delaware supply were estimated at billions of dollars to build and hundreds of millions a year to run (National Research Council, 2000). The city chose to pay to protect the watershed instead.

That is a market with a buyer on it, and it is rare. Many headwater forests have dependents and no market at all. There is no account that belongs to the place, no price, no seller of record, and no buyer. Both sides are empty. The people downstream still depend on it; they just have nowhere to pay and nothing to buy.

:::johnson
**someone has to go first.** *Every market starts empty. Someone goes first, in public, or nobody goes at all.*

[see the agents that go first →](/agents?from=guide)
:::

## how marketplaces get past it

### how do marketplaces solve chicken-and-egg?

They fill one side by hand. Paul Graham's 2013 essay *Do Things that Don't Scale* tells how Airbnb's founders went door to door in New York, recruiting hosts and helping them improve their listings. Stripe's founders did not send a link to someone who agreed to try the product; they set it up on that person's laptop on the spot. Graham calls that a "Collison installation."

The tactics differ. The shape is the same:

| move | what it does |
|---|---|
| recruit one side by hand | gives the other side something real to arrive into |
| pay the harder side to show up | buys time until the market pays them on its own |
| be the first user yourself | puts a real participant on a side, at your own cost |

Many marketplaces got past their cold start because someone filled one side by hand. The question is never whether to seed. It is how to seed without lying about it.

## seeding is not faking

There are two ways to make an empty market look full, and only one of them is a market.

**Seeding** puts a real participant on one side, under its own name, with a reason anyone can read. The founder who lists the first room is a real host. The protocol that buys the first certificate is a real buyer, and the money really moves.

**Faking** puts an appearance on a side: made-up accounts, listings nobody stands behind, or a wallet trading with itself. CME Group's market regulation material defines a **wash trade** as a transaction, or series of them, that gives the appearance of authentic purchases and sales without a bona fide market position. Nobody's position changes. The only product is the look of demand.

| | seeding | faking |
|---|---|---|
| who is on the other side | a named participant | nobody, or yourself |
| is it disclosed | yes, by name | no, that is the point |
| does anyone's position change | yes | no |
| what a stranger can check | the account, the reason, the trade | a number with nothing under it |

The line we hold is simple. Every trade an agent makes has a public mandate that explains it, and an agent is never its own counterparty: no wallet buys from itself to make a number. If a trade cannot be explained by the mandate of the account that made it, it is not evidence of anything, and we do not point to it.

## how this market starts

Here is the cold start for nature, stated plainly.

More than 2,000 agents exist on [*ensurance*](/agents?from=guide). An **agent** is an account for a place, a people, or a purpose: a watershed, a land trust, a species. Each one is a token that carries its own wallet, so money can be held for that place and moved on its behalf, separate from anyone's personal funds. Is this crypto? Yes: the accounts sit on Base, a public blockchain, which is why anyone can read every trade they make.

The protocol funds agents first and trades through them first. That is the order, and it is public:

1. **Account for the place.** The agent exists before any outside buyer does.
2. **Mint the instruments.** [Certificates](/specific?from=guide) tie a purchase to a named place and pay the people on its split. Coins give the place a market anyone can trade.
3. **Buy first, and say so.** The protocol and its own operators, such as [`tmo.earth`](/tmo.earth?from=guide), are designed to buy first, under their own names.
4. **Let the next buyer arrive into a price** instead of a pitch.

Programs that exist let an agent accumulate a target coin or spread its holdings across coins, inside a mandate its owner turned on. In the agent's chat, [the account proposes and a person confirms](/guide/the-account-proposes-you-confirm?from=guide). A scheduled purchase of a steward's certificate by the place that depends on them is planned. Each of them is meant to sit inside the line above: named, explained, never with itself.

### is it honest for a protocol to be its own first user?

Yes, when three things are true: the protocol is named as the buyer, the trade follows a mandate anyone can read, and no account trades with itself. Under those conditions, a protocol buying first is the same move as a founder listing the first room. Without them, it is a painted market.

Being our own first user is the design, not a fallback. A forest does not wait for the water cycle to be perfect before it grows; its canopy helps recycle the moisture it needs. [The permanent bid](/guide/the-permanent-bid?from=guide) covers that pattern and the flywheel behind it. This post only claims the first step: somebody has to stand on the empty side, and it is us, in the open.

## pipe or payor

Not all first money is the same, so we say which kind it is every time. When protocol money buys a steward's certificate, that is a **pipe**: seed money moving from inside the system to the people doing the work. It pays them. It does not prove anyone outside needs them. It becomes a **payor** when a dependent's own balance sheet refills the account, the way New York City paid for the Catskills because the plant it would otherwise have had to build was estimated to cost far more. A pipe starts a market. A payor is the market. Both are disclosed, and neither is dressed up as the other.

## what this means for you

**If you run an onchain group,** you can stand up an agent for a place your members care about and be its first buyer, by name. Your group's purchase is seeding, not faking, as long as it is public and explained. The [onchain groups page](/solutions/onchain-groups?from=guide&topic=the-cavalry-is-not-coming) sets out what that looks like.

**If you build or run an AI agent,** an ensurance agent is a counterparty with a readable mandate. You can see why it holds what it holds before you trade with it. The [AI agents page](/solutions/ai-agents?from=guide&topic=the-cavalry-is-not-coming) covers the surface.

**If you provide capital,** you do not have to be the first buyer. The protocol is designed to be. You can invest into a market that has an account, an instrument, and a named first participant. Any return comes from what the instrument does; no figure is promised here. For the current numbers, read [the $ENSURE page](/ensure?from=guide) and [the pools](/pools?from=guide) yourself.

This post is part of [the cavalry is not coming](/guide/the-cavalry-is-not-coming?from=guide), a series on how capital forms for nature from inside when no outside check is on its way.

Every market starts empty. This one starts with us on the empty side, buying first from a place's own account, by name and inside a public mandate, so the next buyer is not the first.

[browse the agents →](/agents?from=guide) · [see how onchain groups take part →](/solutions/onchain-groups?from=guide&topic=the-cavalry-is-not-coming) · [talk to someone about seeding a place →](/contact?from=guide&topic=the-cavalry-is-not-coming)

## sources

[Caillaud, B. and Jullien, B. (2003), "Chicken & Egg: Competition among Intermediation Service Providers," RAND Journal of Economics 34(2)](https://ideas.repec.org/a/rje/randje/v34y2003i2p309-28.html) — the two-sided market trap named in a title

[Andrew Chen, *The Cold Start Problem* (Harper Business, 2021)](https://a16z.com/books/the-cold-start-problem/) — network effects and why a new network starts with little value

[Paul Graham, "Do Things that Don't Scale" (2013)](https://paulgraham.com/ds.html) — Airbnb going door to door in New York; the "Collison installation"

[CME Group, "Definition of a wash trade"](https://www.cmegroup.com/education/courses/market-regulation/wash-trades/definition-of-a-wash-trade.html) — wash trading as the appearance of purchases and sales without a bona fide position

[National Research Council (2000), *Watershed Management for Potable Water Supply: Assessing the New York City Strategy*](https://nap.nationalacademies.org/catalog/9677/watershed-management-for-potable-water-supply-assessing-the-new-york) — nine million people served; the city's filtration cost estimate

## the series

[the cavalry is not coming](/guide/the-cavalry-is-not-coming?from=guide) (start here)

1. [every market starts empty](/guide/every-market-starts-empty?from=guide)
2. [the first check](/guide/the-first-check?from=guide)
3. [who stands on the other side](/guide/who-stands-on-the-other-side?from=guide)
4. [what turns the flywheel](/guide/what-turns-the-flywheel?from=guide)
5. [the first print](/guide/the-first-print?from=guide)
6. [why the second buyer pays more](/guide/why-the-second-buyer-pays-more?from=guide)