A community with an economy attached.
Everyone who helps the platform grow gets paid for it, and their earnings can pay for their own use. It is one economy with several ways in, built on the same provenance ledger that measures the cross-tenant cache. That is what lets every earning be self-policing: a builder earns only when their work is actually run, a referrer only when the referred user actually pays, a seeder only when others actually reuse their verified result.
- Live at launch
- Build (the builder royalty) and Refer (the ring-fenced acquisition layer).
- Staged
- Seed (the first-computer reward), turned on when cross-tenant reuse density warrants.
- The connective layer
- Earn-to-use: any earning can offset the earner's own bill.
Communities do not fork.
The evidence that opened this raise is that ten paying customers are inviting their friends, unprompted, and staying for the community as much as the tool. That is the supply side of a flywheel showing up on its own, the hardest side to manufacture. The participation economy is the structure that turns that organic community into a compounding, defensible asset.
| Community without an economy | Community with an economy attached |
|---|---|
| Members stay for the product | Members stay for the product, the people, and the income |
| A better tool can lure them away | Leaving means abandoning earnings and published work that keeps paying |
| Supply of capability is the vendor's roadmap | Supply of capability grows on its own as builders earn |
| Copyable by a vendor with more capital | Not copyable; a vendor cannot pay an open builder base against itself |
A model vendor can copy a feature. It cannot copy a community with an economy attached, because communities do not fork. When a member earns from building, referring, and seeding, and when those earnings can offset their own bill, leaving is not switching a tool; it is walking away from an income stream and a place where their published work keeps paying them.
One economy, several ways in.
There are four earning motions. Two are live at launch, one turns on as the network matures, and one is the way any earning can offset a bill. Earn-to-use is the connective idea that makes the economy feel like a community rather than a marketplace: a member's contributions can pay for their own consumption.
| Motion | What you do | What you earn | Status |
|---|---|---|---|
| Build | Author an action | A royalty every time anyone runs it | Live at launch |
| Refer | Bring someone to the platform | A declining share of what they pay | Live at launch |
| Seed | Produce a verified result others inherit | A reward each time others reuse your verified work | Roadmap; turns on when reuse density warrants |
| Earn-to-use | Roll any of the above against your own bill | Your usage offset by your earnings | The connective layer |
Metering is individual; payment is flexible.
The economy separates who generates usage from who pays for it, and that decoupling is the adoption unlock. Every person's usage is metered per person, always. That is the atomic unit and it never moves. Who pays is a separate, flexible layer, which removes the everyone-needs-their-own-card friction.
| Payer mode | Description |
|---|---|
| Self-pay | The default; a person pays for their own usage |
| Team-absorb | An admin or operator picks up some or all of their people's usage |
| Sponsor | A third party covers a person's usage |
| Prepaid credits | Usage drawn from a pre-purchased pool |
| Earn-to-use | Earnings from build, refer, or seed offset the bill |
The more an operator absorbs, the more the shared cache saves across their people, because their people run overlapping work on the shared surface. The operator motion and the cache flywheel are the same motion, which is why the operator and agency channel is the natural way for Quinn to travel.
Less on tokens every year, on purpose.
The economy is honest about where money comes from today and where it comes from as the platform matures. Every cache hit is positive-sum in the near term: the payer pays a fraction of fresh-compute cost, the builder earns, and Light Brands keeps a margin. But the durable business lives in the value that grows with the network, not in the compute spread that shrinks with compute.
| Durable revenue line | What it is | Why it is durable |
|---|---|---|
| Certainty | Proven-correct, gate-verified results | People pay for a guarantee whose value is independent of compute cost; the one line that does not deflate, so it is led with and built first |
| The standard | Being the layer everything runs on | A standard captures value an application cannot |
| Registry take-rate | A cut of marketplace value | Scales with the network as builder supply grows |
| Access | The usage bundle, a flat fee for being on the platform | Recurring, predictable |
This is a deliberate inversion of the usual AI-application model, which depends on a margin between price and token cost that compresses as compute cheapens. Quinn plans for that compression by design and moves its revenue to lines that grow with the network.
The builder royalty.
Building an action is the supply-side motion of the flywheel, live at launch. Because a builder is paid every time their published work runs, the supply of capability grows on its own: builders have a standing incentive to author, maintain, and improve actions, and every action they publish adds gravity to the platform.
| Property | Detail |
|---|---|
| What earns | An action a builder authored, every time anyone runs it |
| Pricing basis | The outcome price is a fraction of recompute cost, not a token markup, which is what lets the royalty hold up as compute trends to free |
| Self-policing | The builder earns only when their action is actually run; demand cannot be faked once the trust layer is enforced |
| Portability | An action works across accounts, so a single published action can earn across many customers and many operators' client books |
The saving and the payout come out of the same event, which is unusual: on a cache hit, the buyer's cost falls and the builder earns, from the single act of reuse. That is what could turn a marketplace into one that sustains itself.
The ring-fenced acquisition layer.
The referral is different in kind from the work-value earnings, and it is deliberately kept architecturally separate from the standard. It is a Light Brands company growth mechanism, not a rule of the registry. Referred users and walk-in users get provably identical metering, pricing, and cache economics, and that identity is provable from the ledger.
| Property | Detail |
|---|---|
| What earns | Bringing a new user who pays |
| The shape | A declining share of the referred user's payments, single-level, funded entirely from Light Brands' margin |
| Declining, not perpetual | Full rate for an initial window, declining to a durable floor that is never zero; the referrer is honored permanently at the floor |
| Single-level | One hop only; no multi-level chains, ever |
| Self-liquidating | Paid only when the referred user actually pays; acquisition cost is a fraction of realized revenue |
| Versioned, no-clawback | Every referral bond is honored on its terms; the rate for new referrals changes only forward |
The operator wedge: three earnings stacked on one motion
The operator and agency channel is the deepest leverage in the economy because a single operator motion earns three ways at once, all from the same act of bringing and serving a client book. It is the offer the static-workflow incumbents cannot write, because they have neither the cross-tenant cache nor the per-run builder economy to stack on the referral.
Referral
Deploy-many
Cache warming
The referral is self-liquidating: a dollar of acquisition cost is only ever spent as a fraction of a dollar the referred user actually paid. Light Brands never spends acquiring a customer who generates no revenue. The single-level, declining-to-a-floor shape kills the whale problem and keeps the standard credibly neutral, because the referral lives beside the standard, not inside it.
The staged first-computer reward.
The seeder motion is designed and reserved, not live at launch. It turns on when cross-tenant reuse is dense enough that dividing it is non-trivial. Rewarding first-computation only makes sense when reuse is dense enough to divide, and reuse density is exactly what is being measured now. Measure the secret before monetizing it.
| Property | Detail |
|---|---|
| What earns | Being the first-computer: paying to produce a verified result that others later inherit |
| The reward | A share each time the seeded result is reused |
| Decay by reuse | A viral result graduates to the commons fast; the seeder's share decays toward a perpetual floor |
| Decay is a transfer, not a loss | The decayed portion transfers to the builder who maintains the action; the contributor pool stays constant |
| Claimable authorship | An abandoned-but-reused action can be adopted by an active member with approval; the original author keeps a perpetual created-by credit |
The seeder economy is where the no-clawback guarantee is demonstrated most clearly. Turning it on obeys no-clawback: the seeder's share is funded from Light Brands' margin, never by cutting the builder. A new way to earn appears, and no existing earner is reduced to fund it. The staging itself is a demonstration of the guarantee.
A payout system without a fraud model is a bug bounty for your own treasury.
Every earning is self-policing only if the platform knows who is who and whether value is real. The moment money is in the ledger, adversaries have a direct incentive to fake it. The trust layer is first-class and ships with the economics, on the same provenance ledger, with the same rigor as the cache proof. This is a gate before the economy goes live, not a version-two concern.
| Defense | What it prevents |
|---|---|
| Distinct-payer diversity | A dividend accrues only above a minimum number of genuinely distinct, genuinely paying reusers, so a ring cannot manufacture demand among sock-puppets |
| Sybil and self-dealing defense | Identity resolution, no self-referral, detection of operator-refers-own-absorbed-team loops, and recycled-account checks |
| Attribution rules, written down | What counts as a genuinely new referred user, the attribution window, and last-touch resolution when several operators touched a user |
| A red-team memo that survives a hostile read | Before a single dividend is paid: model the colluding ring, the wash-trade, and the front-run, and show the ledger catches each |
Never claw back. Here is the receipt.
The economy is governed so that betrayal is impossible by design, not by promise. This is what makes it a standard rather than a company's changeable pricing page. What has already been earned or promised is never reduced; the policy changes only forward, and new earners are funded from Light Brands' own share, never by cutting an existing one.
| Governance property | How it is made credible |
|---|---|
| Version the economics | Every action and every referral binds to a policy version; new versions apply only forward |
| Stamp the policy version on the ledger | Every earning carries the version it was made under |
| Publish a hash of each policy version to an append-only log | The difference between we promise we will not claw back and we cannot claw back, here is the receipt |
| Prove neutrality from the ledger | Referred and walk-in users can be shown to receive identical standard economics |
| Commodity escape valve | A forward-only policy lever on marginal new activity, never a claw-back |
The economics go live behind two gates, neither of which is tuning
Trust
The curve
The one thing a vendor with more capital still cannot buy.
The participation economy is the defensibility asset a well-capitalized model vendor cannot replicate, and the reasons are structural. Every part reinforces every other part, and the whole is a community with an economy attached.
| Requirement to copy the economy | The model vendor's position |
|---|---|
| Pay an open builder base per run | A single vendor optimizing its own margin cannot credibly pay an open builder base against itself |
| Fund new earners from margin without clawing back | Requires the durable-layer revenue model and the no-clawback governance, which are a strategy, not a feature |
| A community that stays for the income and the people | Communities do not fork; a vendor cannot buy an incumbent's community |
| Prove neutrality from an append-only ledger | Requires the ring-fenced referral and the hashed policy log; a vendor's growth mechanics are not neutral by design |
The cache makes each run positive-sum, so building and seeding pay. The builder economy makes the supply of capability grow on its own, which is the library-density that gives the open standard its gravity. The referral is negative-risk CAC that supercharges the operator wedge, and the operator wedge is the same motion that warms the cache. A model vendor can copy a feature. It cannot copy a community with an economy attached.