Frequently Asked Questions
Short answers to the questions people ask most about TensorSwing. For the full mechanics, see Methodology; for what's in and out of the fixed universe, see The Universe Rule.
Most questions about an AI stock picker come down to three things: how accurate it is, whether the accuracy claim can be trusted, and whether it touches your money. TensorSwing's answers are that the track record is published in full including losing weeks, that every call is hash-committed before the week starts so it can be verified without trusting us, and that the app never connects to a brokerage or places a trade.
01 What is TensorSwing?
An iOS app where seven forecasting models — two trained in-house and five open foundation models — each call the five-day move of every one of 119 US stocks and ETFs, every Sunday. Every call is cryptographically committed before Monday's market open, its outcome labeled Friday, and a sample spot-checked weekly by a public randomness beacon the app's operator cannot influence. Every week stays in the record. It is informational research commentary, not investment advice, and no trades are placed in the app.
02 How does the weekly verdict process work?
Sunday evening, every model calls every name in the fixed universe and all 833 calls are hashed and sealed before Monday's open. The call stays sealed all week — tracked as ahead or behind, never win or loss. Friday afternoon it's graded against the closing price and published as win, loss, or scratch. Saturday, a public randomness beacon selects a sample — plus every scratch — to open and check against its hash. Full walkthrough: Methodology.
03 What does verification actually prove — and not prove?
It proves the committed calls existed, unedited, before their outcomes were known; that the sealed set can't be silently added to or pruned after the fact; and that the beacon-selected sample's revealed calls match their published labels. It does not prove full-set accuracy, model quality, or profitability, and it says nothing about future weeks.
04 What's the difference between Free and Premium?
Free reads the house models' live calls on every stock, keeps four names on a watchlist with one alert, and sees the last eight graded weeks. Premium ($39/month) reads all seven models, keeps twenty names and five alerts, and unlocks the full archive, the proof pages and the per-stock AI news reads. The graded record is never gated. There is no trial.
05 Does TensorSwing place trades or send trading alerts?
No. TensorSwing is informational only — it does not execute, route, or facilitate any transaction, and it is not a broker-dealer or registered investment adviser. It also does not send accuracy or streak push notifications; that class of alert is deliberately not built. Full risk language: Risk Disclosure.
06 How is the universe of tradeable names decided?
By a public, computable rule anyone can run against public data — not by the models' own judgment. It selects US-listed common stocks above a volume and market-cap floor, plus a fixed list of ETFs, and recomputes quarterly. Full rule: The Universe Rule.
07 Can I verify a week myself, without trusting TensorSwing?
Yes. Every manifest is anchored to the Bitcoin blockchain via OpenTimestamps, and the record, protocol specification, and verification tooling are published at github.com/YermekIbrayev/tensorswing-record.
08 Still have a question?
Reach out via the Contact page.