# Verifiable Forecast Log: legal boundary note (written BEFORE any code)

Date: 2026-11-14 (GMT+2). Status: internal design note by an AI agent acting for QANTS. **This is not legal advice and no lawyer has reviewed it.** It sets the product's boundary so the build cannot drift into a regulated activity. Anything marked "needs counsel" must be reviewed by a qualified lawyer before the product is opened beyond a free, educational testnet-style demo, and certainly before any real money is charged.

## 1. What the product is
A tamper-evident diary of **probabilistic predictions about non-financial, publicly checkable events**. Each entry is hash-chained to the previous one; the chain head can be timestamped on Bitcoin. After the event, the author records the outcome with a public source. The report scores the author's stated probabilities (Brier score, calibration table) with the sample size shown.

## 2. What it must never become
1. **A signals or advice channel.** No entry about the price, direction, level, yield or return of any tradable instrument (shares, ETFs, indices, futures, options, currencies, crypto-assets, tokens, commodities, rates). No "buy/sell/long/short/target" language. No entries that a reader could act on as a trade.
2. **A paid tips service.** No fees for access to forecasts, no subscriptions to a forecaster's picks, no pay-per-call, no tips, no affiliate links tied to a call. Many regulators treat paid, specific, actionable calls on instruments as regulated activity; a log of weather or sports outcomes sold as "picks" can drift into gambling or betting-advice rules. **Needs counsel** before any charging.
3. **A betting or prediction market.** No collateral, no wagering, no payouts, no odds, no prizes. tQNT, a testnet unit with no cash value, is not used by this product at all.
4. **A reputation weapon.** No predictions about named private individuals, no allegations, no claims about the conduct of identifiable people or companies (defamation and privacy risk). Entries are about events (weather, scheduled releases, sports results, measurements, public chain statistics).
5. **A cherry-picking machine.** A "track record" that shows only the wins is misleading. The chain covers every entry; the report always shows how many entries are unresolved and how many committed entries were never revealed.

## 3. Rules enforced in code (and re-checked by the verifier, so a modified tool cannot hide a violation)
- **Category allow-list** (anything else is refused): `weather`, `sports_result`, `scheduled_release`, `science_observation`, `public_chain_data`.
- **Deny-list** on the question text: market and instrument vocabulary (price, stock, share, ETF, futures, options, forex, crypto, token, yield, earnings, IPO, buy, sell, long, short, bullish, bearish, rally, crash, portfolio, hedge, ticker symbols like `$ABC`...). It is deliberately strict and will refuse some harmless text; there is **no override flag**.
- A **public resolution source** (a URL) is required at entry time, fixed before the outcome, so the author cannot pick the source afterwards.
- Probabilities are whole basis points from 1% to 99%. No certainty claims.
- Entries cannot be edited; a correction is a new entry that references the old one.
- Every report carries: "educational; not financial advice; not a signal; measures stated probabilities against outcomes on this log only".

## 4. Data and privacy
- The public chain contains only what the author chooses to publish. **Committed mode** publishes only a commitment hash, the category and the resolve-by date until reveal; plaintext can stay private. Authors should be told that a hash chain **cannot be erased**: do not put personal data in an entry. Revealed text can be withdrawn from a website but the hash remains.
- QANTS stores nothing in this MVP: the tool is local, the sample is synthetic.
- Any hosted version would need a privacy notice, a removal policy for plaintext, and age and jurisdiction checks. **Needs counsel.**

## 5. Disputes and integrity
- Outcomes come from the public source named at entry time. If the source is unavailable or ambiguous, the entry is `void` (counted separately, never as a win or a loss).
- Self-reporting is the main weakness: the author records the outcome. The log proves **order and non-editing only for entries covered by a head that was stamped or published** (anything after that head could be rewritten), and not that the author reported honestly. A third-party resolver would be a later feature.
- One person can run many logs and publish only the best. A single log is not a ranking.

## 6. Jurisdiction and money
- The brand is global. No jurisdiction is assumed. Rules about advice, betting and promotions differ widely; the safe default is the narrowest reading above.
- **No real-money charging until a legal entity exists and counsel has reviewed this note.** Until then the product is a free, educational tool with a waitlist release.

## 7. Launch gate
Free local tool + synthetic public demo: allowed now. Hosted submissions from the public, any fee, any ranking or leaderboard of forecasters, any partnership with analysts or newsletters: **blocked pending counsel and QANTS' explicit sign-off.**