One token per concept. A price that is a formula, not a person. A way to sell. And a threshold at which information can follow. These are the rules for the OuterTalk map — written down before they are built, so that serious people can tell us where they are wrong.
That sentence is the whole thesis, and it needs one thing to be honest: a way for money to be wrong and leave.
Every category on the OuterTalk map — and every concept under one — has its own token. You buy it because you believe in the concept, whether or not anyone official does. If enough people back it, it lights, and the machinery that produces information turns on for it: posts, polls, a page, an assistant that will draft there.
If the concept turns out to be wrong, holders sell. Some hold on. The price falls because people left, not because anyone decreed it. A token that can fall is the honest cost of backing something that turns out not to be true. That is what makes this a record of belief rather than a popularity contest with no exit.
None of this decides what is true. A separate record, kept by people whose money counts for nothing in it, does that. Price and truth never touch. That rule has a name here — the wall — and it is the one thing on this page we will not move.
The loop · money on the left, meaning on the right, one wall between them
Opens, takes stakes, closes, pays out to whichever side carried more. No withdrawals. It is a consensus instrument, and it is not what this page is about — it is a separate thing with its own rules.
No bets, no settlement, no clock. Back a concept, hold, sell when you change your mind. This page is the rules for layer 2.
Two instruments · the difference is a clock and an exit
Each rule says whether it is decided, still open, or decided-but-unbuilt. Where a rule reverses something we wrote earlier, it says so.
Every place on the map — one of the 42 categories, or a concept under one, like Frequency Healing under Health — has its own token. Buying it says “I back this.” It buys visibility and, once the concept lights, the right to open a Smart Poll there.
It never buys truth, a name, a vote, or a position on the map. A token belongs to its address on the map, not to the idea currently sitting there; if the idea is renamed or retired, the token stays.
Anyone signed in can propose a concept: pick one parent, give it a name, and make the first backing (at least 100 ✦). Nobody is handed tokens for free — the founder backs at the same price as everyone.
Proposing and buying are two separate actions: the name and parent are a matter of meaning and cost nothing; the backing is money. The name is provisional until people review it. It must be neutral — no “Fake”, “Pseudoscience”, “Big X” — and cannot exactly match an existing one. You may have at most two unlit proposals at a time.
Concepts live in the category tree, under a parent. Topics are something else: they connect things across categories, and they are not tokens.
It starts at 1 ✦ and rises by 1 ✦ for every 2,000 tokens in circulation. Buying pushes it up; selling pushes it down. Nobody sets it by hand and nothing else moves it — not a vote, not a verdict, not us.
After a concept lights, any holder can sell back to the token's own reserve — the pool of every credit ever paid in. There is no other side to the trade: no buyer needs to exist. The reserve is always enough to pay every holder out at the formula price, in any order, and by design it is never withdrawn, lent or seized. Today that is a database row we administer and a test we publish; on chain it becomes a reserve no key can reach.
A 2% exit fee is burned. It goes to no one — not to the reserve, not to other holders, not to us.
Every ✦ a seller takes out was put in by another backer. There is no yield, no dividend, no fee share, no reward for holding. If you sell for more than you paid, it is because someone backed the concept after you at a higher price. Credits return to your OuterTalk balance and nowhere else.
A concept lights when two floors are met at once: at least five different accounts each holding tokens they bought at least 72 hours ago, and a reserve of at least 1,000 ✦. Deeper concepts light at three accounts and 500 ✦.
Both floors, on purpose: the people floor stops one wallet from buying a concept outright; the money floor stops five friends from carpeting the map for pocket change; the 72-hour age stops anyone from buying, lighting, and leaving the same day. Lighting changes the price by nothing and says nothing about whether the concept is true.
The trigger is money and people — not a verdict. Nobody has to approve a concept for it to light. What people do approve is its name (R2) and, separately, what the evidence says about it (R9).
Rule 6 · why both floors · four ways to put money on a concept
Backing on an unlit proposal is locked. You cannot sell it. It sits there accumulating toward the two floors, and the parent shows only a count — “1 proposal pending under Health” — with no name and no spot on the map.
A proposal has a window to light. If it never reaches both floors, every backer is returned exactly what they paid, with no fee, and the proposal is withdrawn as if it had never existed. Nothing is consumed and nothing is left behind. The window is 30 days, and it restarts whenever new backing arrives: a proposal that keeps drawing people stays alive; one that stalls is returned.
A concept gets its permanent spot on the map — and its name rendered there — the day it lights, after its name and parent have been reviewed. Not the day it is proposed. Once lit, the spot is fixed forever: nothing on the map is ever deleted or moved.
A lit token whose reserve later falls to zero is a dormant ember: still there, still buyable, still sellable, and it re-lights the moment it re-crosses the floors.
A concept is disproven only when the evidence record says so — decided by people whose money counts for zero in that decision. A falling price never disproves anything, and a disproof never freezes buying, forces a sale, or changes a price.
What a settled disproof does is put a disclosure on the buy screen: the claim, the date, a link. You can still hold, or sell, or both. If holders leave, the price falls because they left.
No account may hold more than 25% of a token's supply, or 500 tokens — whichever is larger. A buy over the line is refused with the allowed amount shown. Every purchase is held for 24 hours before it can be sold.
How brightly a concept glows on the map is set by what happens in it — the volume of real polls and posts, netted and fading over 24 hours — not by how much money is in its reserve. Buying and selling tokens never counts as activity and never moves brightness. There is nothing to wash.
The reserve, price and holder count are shown on every token's panel (R12). They are facts about money, and they are displayed as facts, not as glow.
Rule 11 · sixty hours in the life of one concept
Every token shows its own panel: reserve, price, supply, holder count, top-holder share, 24-hour net flow, lit or dark, any consensus record — and the sentence “selling pays you from this reserve; nobody else is on the other side.” No price chart is ever shown without it.
Only lit concepts can have children. Child slots open in waves — 1, then 2, then 3, then 5 — each wave opening when the last is fully lit. A child is backed by whoever proposes it; the parent's holders pay nothing and are paid nothing. A proposal that never lights (R7) never occupied a slot, so a wave is never blocked by junk.
Rules 8 and 13 · a lit parent, its child slots, and what a dark proposal does not get
Making accounts is free, and every account starts with 10,000 ✦. So every limit on this page is a speed bump, not a wall. The wall is proof of personhood, which is in design. Until it exists, “five accounts” means five accounts, and we say so rather than pretend otherwise.
What the rules above do in the meantime: locked backing (R7) means lighting costs three days of held money, not a same-day round trip; earning the address at lighting (R8) means a griefer's junk never lands on the map; and a failed proposal is returned whole, so there is nothing to gain by spamming.
Our earlier design gave these tokens no market on purpose: you could lock credits into a concept and that was all. The reason was legal. A token people buy hoping it rises from other people's work is what U.S. securities law — the Howey test — was written to catch, and one you can sell for more than you paid looks more like that than one you can only lock.
We are adding selling anyway, because a token that can fall is the honest cost of backing something that turns out to be wrong. What that means today: we believe the first prong of Howey — money invested — is not met while credits are free, unsellable and uncashable. Nothing here is an offer of anything. If real money ever enters, that is a different product, built after securities counsel, with any fee-to-holder mechanism off until counsel clears it. If you think we have this wrong, tell us.
See the ledger below. Rules 2 through 13 are unbuilt. What exists today is a backing function that only adds, 42 category rows with placeholder scores, and seven money contracts on a testnet.
Two things we wrote down earlier are reversed here on purpose: “lock-only, no trading curve” (June 2026) and “no AMM — an AMM would manufacture a security” (September 2026). The reasons they were written are in R15 and are still true. A third document, from later in June, had already marked the lock-only rule as revised; this page finishes that thought.
Also replaced: the simulator's placeholder rule that any child concept with a nonzero score counts as active (it never fired — there are no children), and the placeholder scores themselves, which are a hash of a row id and will be zeroed before the first real backing exists.
No selling. No address. No rendered name — the parent shows a count. Holders and reserve accumulate toward both floors.
Address welded, name shown, picker row, page, posts and polls can be filed, first child wave opens. Selling opens here.
The window closes without both floors met. Every backer gets back exactly what they paid, no fee. Nothing was consumed.
State machine · every arrow is a rule on this page
Integer credits only. Cost rounds up, payout rounds down, and the reserve can never be short.
Worked example, by hand: A backs 500 ✦ at S = 0 → 449 tokens, price 1.22. B backs 500 ✦ → 379 tokens, price 1.41. Reserve 1,000 ✦ ↔ S ≈ 828. One account at its full starting balance moves S to about 4,600 and the price to about 3.3 ✦. These are pinned in a test before anything ships.
price(S) = 1 + S / 2,000 · the reserve is the area under the line
Measured against production on 28 September 2026. Nothing on this page should be read as shipped unless it is in this table.
The path a concept takes · solid is built, dashed is specified and unbuilt
| Thing | State | Detail |
|---|---|---|
| Categories | 42 · built | All top-level. Zero child concepts have ever existed. |
| Backing function | built · adds only | Locks credits into a category. There is no sell, unback or withdraw anywhere. |
| Holdings | 0 rows | Nobody has ever backed a category. |
| Placeholder scores | 42 · synthetic | A hash of each row's id, not money. To be zeroed before the first real backing. |
| Money contracts | 7 · Base Sepolia testnet | Token, treasury, registry, poll factory, poll pool, note staking, resolution. Deployed April 2026 on chain 84532. No mainnet deploy exists. |
| Category token (chain) | built · fixed price | An ordinary transferable token minted at an owner-set price. No curve, no reserve, no redemption. |
| Claim ledger (chain) | built · undeployed | The evidence record R9 depends on. 55 tests pass. Not deployed anywhere. |
| Wallets | 0 of 36 users | The wallet path is switched off in the app. |
| Proof of personhood | in design | See R14. |
| Rules 2 – 13 | unbuilt | Everything above the ledger is a specification. |
Numbered so you can refer to one. Tell us which way, and why. Q1 is decided and kept for reference.
This is not open source. It is open to suggestions, from people who take it seriously. If a rule is gameable, contradictory, or illegal, we would rather hear it now than after it is built.