Picture this

Who it is for

A nurses' community funds a triage-note model. Backers get a share of x402 call revenue and a vote on the next dataset.

Two ways to own it

What you get

Base$399

$100 deposit to reserve

Funding campaign, milestone escrow, x402 endpoint, revenue split to backers

You choose: chain
You choose: name and logo
You choose: accent color
You choose: model domain

Customfrom $3,000

Built around you

Reshape it entirely: new features, your flows, your brand end to end, integrations you need. We scope it with you.

Everything in Base
Custom features and design
Priority build

Why it is new

What exists today, and what this does differently

What is new

Model crowdfunding and revenue-share tokens exist; a turnkey niche fine-tune fund is a narrower packaging.

Closest things out there

Llama Fund crowdfunds open models, Initial Model Offerings and ERC-7641 tokenize models with usage revenue share, and Sentient's OML proposes community-owned models.

Checked with web searches in October 2026. A search is not proof that nothing similar exists, and the market moves fast, so we recheck before every build.

What it unlocks

The human in it

Communities own the AI built from their expertise.

The Regatta Board

Races it can enter

grant · Gitcoin · upcoming

Gitcoin Grants rounds

quarterly-ish rounds; next round dates unknown

Official page ↗

Must be open source

Their rule on existing code: Existing open-source projects OK (generally required)

grant · Optimism Collective · rolling

Optimism Retro Funding

rounds/seasons; current round unknown

Official page ↗

Needs deployed project with measurable impact

Their rule on existing code: Retroactive: existing deployed projects with measurable impact required

What you own

The whole app, not a starter kit.

The complete Fine Tune Fund app: a community crowdfunds a small open AI model into milestone escrow, releases money only when the model passes its own eval, serves it pay per call, and shares usage fees with backers.

Base price · $399

The user app, complete

Every buyer gets the same finished app. Basic branding and copy changes come through a quick form.

Three step onboardingHome

Shown once: your community funds it, it gets paid only when it passes, clinics pay per draft.

Fund status and next actionHome

Raised, backers, milestones, and one clear next step, with the safety rule pinned.

Campaign with refund if missedFund

Goal, progress jar, days left and milestone budget breakdown.

Pledge through checkoutFund

Pledge USDC into escrow through the shared checkout with a receipt.

Milestone evalBuild

Compare the first fine-tune with the corrected one; invented values against the nurses' line.

Reviewer quorum signingBuild

Volunteer reviewers sign, and escrow releases only after quorum and a passing eval.

One tap eval rerunsBuild

Rerun the full eval after each fix. Gated behind Hosted upkeep.

x402 endpoint launchUsage

Pay per draft with a clinic key, usage and backer pool totals.

Per caller capsUsage

Daily cap per key that stops a retry loop draining a clinic wallet.

Spend alerts and quality monitorUsage

Alerts per caller and a weekly quality trend. Gated behind Hosted upkeep.

Dataset vote and backer shareMe

One vote per backer wallet on the next dataset, and the backer's share of weekly splits.

Plans, share card and permissionsMe

Clinic key and Hosted upkeep through checkout, the we own it card, permissions and vision.

Open the full app demo

Add-on

Owner console

Your admin side: revenue, paywall editor, users, rules and the agent log.

Revenue and paywall editorRevenue

Campaign fees, per call cut, upkeep plans, USDC payments, payment methods, live pricing and gates.

RulesRules

Reviewer quorum, max invented values, backer share, cap per caller, refund if missed.

Build log and roadmapAgent

What failed and how it was fixed, running costs, and priced v2 and v3 builds.

Top-up

Growth and referrals

Referral and growth mechanics, switched on when you are ready to grow.

Growth loopsGrowth

Case study webinars, we own it cards, clinic pull and reviewer invites.

The money path, built in

How it earns for you

The owner earns a fee only on successful campaigns, a small cut of every paid call, and monthly upkeep from communities without engineers.

Campaign fee3% of funds raised, only if fundedCampaign page, milestone escrow, eval and reviewer signing, x402 endpoint, backer split, votes

Clinic key (per call)$0.002 to $0.01 per draft, $25 prepaidAPI key for a clinic app, pay per draft over x402 with a daily cap

Hosted upkeep$99 a monthManaged hosting, one tap eval reruns, live quality monitor, spend alerts

Platform cut of about 8 percent on x402 call revenue, set by the owner. Weekly backer splits cost cents in gas.

Prices are yours to set. Illustrative, not a promise of income.

Growth loops, built in

How it spreads

Every funded model becomes a public case study that other communities copy
Backers share the we own it card at launch
Clinics that use one model ask for the next niche
Reviewers bring colleagues into the next eval round

Roadmap · build the next phase with us

Where it goes next

v1 included

The full fund app

Campaign page with refund if missed
Milestone escrow with eval and reviewer quorum
x402 endpoint with per caller caps
Weekly backer split and dataset vote
Plans and checkout

Included

Eval studio

Reviewers build eval sets in the browser
Side by side model comparisons
Automatic de-identification checks
Public eval report pages

Add-on from $1,200

Model marketplace

Directory of community owned models
One key for clinics across models
Revenue split across model versions
Grant matching for public interest models

Add-on from $2,500

For investors and grant reviewers

The pitch, ready

Communities crowdfund small open AI models from their own expertise, pay for them only when they pass their own tests, and share the usage fees.

Problem: Niche professional communities have the know-how to build useful small models, but no way to fund, verify and own them; big labs build generic tools without them.

Solution: Milestone escrow releases funds only after a community eval and reviewer quorum, the model is served pay per call over x402, and a split contract shares usage with backers.

Why now: LoRA fine-tunes of small open models cost tens of dollars, x402 makes per call payment simple, and stablecoin escrow makes community funding safe to refund.

Market: Professional communities such as nurses, midwives, paralegals and translators, and the clinics and apps that would pay per call.

The ask: Pilot with two nursing communities and five clinics, plus grant support from public goods programmes on Base and Arbitrum.

Funding is never guaranteed. Check each program's rules.

For hackathons and demo days

Your 2 minute demo

Home: meet Marisol, a charge nurse whose 3,400 member group wants a handover drafting model.
Fund: back the model with a $20 pledge through the USDC checkout.
Build: the first fine-tune invented a blood pressure; escrow holds.
Build: switch to the corrected run, watch the eval pass, and collect reviewer signatures to release milestone 2.
Usage: launch the endpoint, buy a clinic key, try a draft, and switch on caps for a runaway caller.
Me: vote on the next dataset and see the backer share.
Owner Revenue: change the campaign fee and per draft price, and preview the Plans sheet.
Owner Growth and Roadmap: run a case study webinar, then request the eval studio.

Many hackathons require code written at the event. See the Regatta Boardfor each race's rule.

Make it entirely yours

Custom work from $3,000: your community's eval design, a specific base model or host, integration with a clinic system, or a branded fund platform. Scoped on a call.

Custom from $3,000

BLUEPRINT

Why it matters and how it really works

The need

Communities with deep shared know-how, like nurses, farmers or court interpreters, are poorly served by general AI tools, and cannot safely paste their work into them. They could fund a small model that fits their niche, but have no simple way to pool money, check the work and share in what it earns.

What exists and what is new

Model crowdfunding and revenue share already exist: Llama Fund crowdfunds open models, Initial Model Offerings and ERC-7641 tokenize models with usage revenue share, and Sentient's OML proposes community owned models. Research verdict: rare, medium novelty. What is different here is narrower packaging: a turnkey fund for one niche fine-tune, with milestone escrow tied to the community's own eval, an x402 pay per call endpoint, and a monthly dataset vote.

The data it runs on

Pledges, the milestone plan, de-identified training and eval sets held off chain, eval results and reviewer signatures, x402 call receipts, and the backer revenue split ledger. No personal or patient data is ever onchain.

1.The mission

When a community's expertise trains an AI, the community funds it, tests it, governs it and shares in it.

2.Why people can't resist it

Money moves only on a pass Backers see each milestone held in escrow until their own reviewers sign the eval.

Built from our words A model that finally reads the group's shorthand, in their languages, is a point of pride worth sharing.

The monthly vote Backers pick the next dataset, and contributors of reviewed data get more weight.

3.How it works for real

Wallets and pledges: Privy embedded wallets on Base or Arbitrum; pledges in USDC into a campaign contract that refunds everyone if the goal is missed.

Milestone escrow: Funds split into milestones (dataset, fine-tune, endpoint, upkeep). Each releases only after an eval result is attested and a reviewer quorum signs (EAS on Base, or a Safe multisig).

Training: A small open model (for example an 8B Llama or Qwen class model) fine-tuned with LoRA on rented GPUs; the community writes and holds the eval set.

Endpoint: An x402 endpoint: callers get HTTP 402, pay per draft in USDC, receive the output. Per caller daily caps and spend alerts.

Revenue split: A split contract sends a set share of fees to the backer pool, distributed pro rata weekly, and the rest to hosting and maintainers.

Governance: Snapshot style votes weighted by pledge and contributed reviewed data choose the next dataset.

4.Cost per use and pricing

A LoRA fine-tune of an 8B model on a few thousand examples takes a few GPU hours, roughly $20 to $100 per run on rented GPUs, plus the eval runs. Serving on a shared GPU or a hosted inference provider costs roughly $0.0002 to $0.001 per short draft. Base gas for weekly batched splits is cents. Example month at 300,000 drafts at $0.004: about $1,200 revenue, about $250 inference, $100 hosting and indexing, the backer share at 40% is about $480.

Campaign fee 3% of funds raised

The app owner takes a fee on successful campaigns only.

Per call $0.002 to $0.01 per draft

Clinics or apps pay per use over x402; the owner keeps a small platform cut.

Hosted upkeep $99 a month

Managed hosting, eval reruns and monitoring for communities without engineers.

5.Growth loop

Each successful fund becomes a public case study with its eval results and usage, which other communities copy. Backers share the we own it card. Clinics that use one model ask for the next niche.

6.Risks and guardrails

Securities and fundraising law: Revenue share to backers can be treated as a security in many countries. Get legal advice, consider donation or membership models, cap amounts, and never market it as an investment.

Patient data: Training on clinical notes requires strict de-identification and consent, and must follow laws like the Philippines Data Privacy Act and HIPAA where relevant. No patient data onchain.

Clinical safety: The model drafts notes for a nurse to check; it must never make clinical decisions. Evals check for invented values, and outputs are labelled as drafts.

Model quality and drift: A fine-tune can fail or degrade. Milestone escrow, reruns of the eval on each update, and a refund vote if a milestone cannot be met.

Usage is not guaranteed: If no one calls the model, backers receive nothing. The app must say so plainly.

Runaway callers: Bugs can send huge volumes. Per caller caps and spend alerts protect both sides.

7.Build order

Campaign page with goal, refund if missed, and USDC pledges
Milestone escrow with eval attestation and reviewer quorum
Eval runner and reviewer signing flow
x402 endpoint with per caller caps
Revenue split contract with weekly batched payouts
Monthly dataset vote and backer dashboard

8.Where this is going

Now
One nurses' group funds one small model that drafts their handover notes.
Next
Many niche funds, from farm extension notes to court interpreting, sharing eval tools.
Later
Community models call each other over x402, and fees flow back to every group whose model was used.
The world it builds
The people whose expertise trains AI own a piece of it, instead of giving it away.

Costs and tools are estimates at the time of writing. General information only, not legal or financial advice.

OWNER PLAYBOOK

Own it, then chart your voyage

Nine routes from one vessel. Tap a port to open its chart.

Why own Fine Tune Fund

Fine Tune Fund lets a community pay for an AI model built from its own expertise and share in what the model earns. Whoever runs the fund holds the campaign, the milestone escrow and the x402 endpoint, so owning it makes you the trusted steward rather than a platform taking a cut from outside. The backers, the dataset votes and the revenue record all live under your name and your wallet.

What’s stopping you?

“I can't code or train models.”

The campaign page, milestone escrow, x402 endpoint and revenue split ship ready. Training is done by a model builder you hire with the escrowed funds, released only when milestones are met. Your job is the community and the stewardship.

“What if the model is bad and nobody uses it?”

That can happen. Keep the first model small and narrow, release funds by milestone, publish evaluations before launch, and be honest with backers that usage revenue is not guaranteed.

“Is revenue sharing with backers legal?”

Sharing revenue can be treated as a security in some countries. Get local legal advice before launching a revenue share, and consider starting with a donation style campaign plus governance votes only.

“Is it really mine?”

The escrow and split contracts deploy from your wallet, your brand runs the campaign, and the community relationship is yours.

“What about the data privacy of a niche like nursing?”

Use only data contributors have rights to and consent for, strip personal details, and publish your data policy before raising a single dollar.

The million-dollar ladder

You earn a stewardship fee on funds raised and a share of x402 usage fees, while backers receive their split. More niche funds under your brand multiply revenue. The ladder below is a target, not a promise.

$500

$4,000

$25,000

$85,000

First fund, 10 backers

10 backers

$50 average stewardship fee

First model live with paying callers

200 paying callers

$20 per month owner share

Three niche models live

1,000 paying callers