Called It
Log what you expect to learn before a course, then prove how much you actually grew.
Train an AI agent on your writing so others can query your knowledge and pay you per answer.
Try it live, freeA tax expert publishes her agent. Visitors pay cents per answer, and she reviews flagged answers weekly.
Knowledge upload, agent chat, per-answer USDC billing, review dashboard
Reshape it entirely: new features, your flows, your brand end to end, integrations you need. We scope it with you.
Creator clones exist offchain; settling per answer onchain to the creator is the new combination.
Delphi.ai sells creator AI clones via paid memberships, and x402 enables pay-per-query APIs, but no product found combining them.
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.
User-owned AI; eligibility unknown
Eligibility unknown
The complete Ask My Agent app: an expert trains an agent only on what they published, visitors pay per cited answer, refusals are free, and flagged answers are refunded and reviewed on Sunday.
Every buyer gets the same finished app. Basic branding and copy changes come through a quick form.
Your admin side: revenue, paywall editor, users, rules and the agent log.
Referral and growth mechanics, switched on when you are ready to grow.
The owner keeps a share of each paid answer, lower on paid plans, plus monthly Pro and Practice subscriptions. Refusals are never charged.
Answer share per tier set by the owner. Each answer costs about 2 cents to run plus under 1 cent of gas.
Prices are yours to set. Illustrative, not a promise of income.
Train an AI agent on your published writing so others can ask it and pay you per cited answer.
Problem: Experts answer the same questions for free in comments and DMs, while generic chatbots answer confidently from stale or wrong sources.
Solution: An agent that answers only from the expert's dated, published work, cites every answer, refuses for free below a confidence line, and pays the expert per answer over x402 with a weekly review loop.
Why now: x402 makes cent sized payments native to the web, retrieval with citations is cheap, and embedded wallets let visitors pay without crypto knowledge.
Market: Independent advisors, newsletter writers and small professional firms with a body of public writing.
The ask: Pilot with 30 newsletter experts in tax, immigration and freelancing, and ecosystem grant support.
Funding is never guaranteed. Check each program's rules.
Many hackathons require code written at the event. See the Regatta Boardfor each race's rule.
Custom work from $3,000: your own model host, a firm branded deployment, CRM and booking integrations, or additional languages. Scoped on a call.
Experts who write for free, like tax advisors, nutritionists and immigration paralegals, get buried in DMs asking questions they already answered. They cannot reply to everyone, and the people asking would happily pay a little for a fast, sourced answer.
Creator AI clones already exist offchain: Delphi.ai sells clones of creators through paid memberships. Pay-per-request APIs exist too: the x402 protocol lets an agent or site charge per call in USDC. What we did not find is the combination: an agent trained only on the expert's own published work, charging per answer over x402 with the payment split onchain to the expert, refusing and not charging below a confidence bar, and a weekly human review of flagged answers. Research verdict: rare, medium novelty.
The expert's published posts, public Q&As and talk transcripts (each dated), per-answer source citations, x402 payment receipts, the split ledger, flags and corrections. Visitor questions are stored privately; client files are never ingested.
Let people who share what they know earn from it while they sleep, with their name, their sources and their judgment on every answer.
Every answer cites you Visitors see which post the answer came from and when you wrote it. Credit and trust travel together.
Unsure means free Below your confidence bar it does not answer and does not charge, and the question lands in your queue as a future post idea.
Ten minutes on Sunday Review a few flagged answers with coffee. Each correction makes the agent better and keeps your name clean.
Knowledge upload: Import a newsletter archive, RSS, transcripts or a Q&A export. Each chunk keeps its publish date and URL, and is embedded into a private vector index.
Answer model: Retrieval plus a mid tier model (for example Claude Sonnet 4.5 via OpenRouter) for answers, with a small model (Claude Haiku 4.5) for the confidence check and refusals. Newer sources win, and old ones are labelled.
Pay per answer: x402 on Base: the visitor's wallet or card-funded embedded wallet (Privy) authorises the USDC price; settlement happens only if an answer is delivered.
Split contract: A simple splitter on Base sends the expert's share and the operator share per answer. NEAR or Ritual can host the agent for owners who want user-owned AI infrastructure.
Answer fingerprints: A hash of each answer and its cited sources is attested onchain so corrections and refunds can be traced. Question text stays private.
Review dashboard: Flagged answers, refusals and probing attempts collect in a weekly queue with one-tap correct and refund.
Per answer: about 4,000 input and 400 output tokens on a mid tier model, roughly $0.015 to $0.02, plus under $0.002 for the confidence check and under $0.01 in Base gas for settlement and split. At $0.25 an answer with a 70% creator share, about $0.04 covers costs and operator margin. Example month: 4,000 answers cost about $80 to run and send about $700 to the creator.
Each answer ends with 'Ask more of Lucia's answers' and a link to the post it cited, which brings readers to the newsletter, and the newsletter links back to the agent. Experts share their weekly 'answered while asleep' card.
Regulated advice: Tax, legal and medical answers can be regulated. Answers are labelled general information, cite dated sources, refuse personal case advice, and offer a paid consult with the human instead.
Outdated answers: Rules change. Sources are dated, newer ones win, old ones are flagged, and wrong answers are refunded automatically after review.
Prompt injection and data leaks: Only published material is ingested, so there is no client data to leak. Probing attempts are refused without charge and logged.
Impersonation: Someone could clone another expert's public posts. Agents require verified ownership of the source channel before listing.
Payment disputes: Charges settle only when an answer is delivered, and refunds are one tap from the review queue.
Costs and tools are estimates at the time of writing. General information only, not legal or financial advice.
Nine routes from one vessel. Tap a port to open its chart.
Ask My Agent trains an AI on your own writing so people can ask it questions and pay you per answer, settled in USDC to your wallet. Creator clone platforms own the model, the audience and the billing; with yours, the knowledge, the price and the payments stay with you. Your expertise keeps working while you sleep, and you review what it says in your name.
It will sometimes. That is why the review dashboard exists: users can flag answers, you review flagged ones weekly, and you can add a short correction that the agent uses from then on. Set a clear disclaimer that answers are AI generated from your writing, not personal advice.
You probably have more than you think: posts, threads, talks, emails you sent answering the same question. Around 20,000 words on one topic is a good start. If you have less, record yourself answering your 30 most common questions and upload the transcripts.
People already pay for expert time. A few cents to a dollar for an instant answer grounded in your work is a small ask, but only if your topic is narrow and valuable. Start with a free daily allowance and charge after.
Yes. Your knowledge base, the billing contract and the payout wallet are yours. You can swap the model or move the frontend later without losing your users' payment history.
It depends on the field. Tax, legal and medical topics often have rules about advice. Frame the agent as education from your published work, add a clear disclaimer, and check with a professional body in your country before charging for answers in a regulated area.
Ask My Agent earns per answer from visitors, from monthly unlimited plans for heavy users, and from team licences where a company gives staff access to your expertise. Later you can host agents for other experts. This is a target to aim at, not a promise.
Illustrative stages, not a forecast. Results depend on you and the market; nothing here is guaranteed.
Log what you expect to learn before a course, then prove how much you actually grew.
Pose a hard question, stake a bounty, and pay the crowd by how much each answer moved the final consensus.
Turn what you are learning into spaced repetition decks that prove your streak and mastery onchain.
A group writes one living briefing together and everyone is paid by the lines that survive edits.
Your highlights and annotations on any article become a collectible, tippable layer others can subscribe to.
Keep a private log of what you read and watch, then sell anonymized slices only when you choose.
Curate a list early, and earn a share when the things you picked go on to be loved.
Teach live in small rooms where students pay per minute and you get paid the second the call ends.