Draft pending fintech-counsel review. The Service makes statements
that resemble investment recommendations (ranked tickers, numerical
inevitability scores). Even framed as "model consensus", that surface
has regulatory implications in some jurisdictions. The operator is
aware of this gap; final language will be set after counsel review.
Until then, treat every output as research output, not as advice.
1. thesignals is not a registered investment adviser
thesignals labs, inc. is not registered as an investment
adviser with the U.S. Securities and Exchange Commission (SEC), the
Financial Industry Regulatory Authority (FINRA), the Commodity
Futures Trading Commission (CFTC), or any equivalent regulator in
any jurisdiction.
Nothing on the Service is personalised to your circumstances. We do
not know your risk tolerance, your time horizon, your tax situation,
your existing portfolio, or your liquidity needs. We are not your
fiduciary.
2. Outputs are not recommendations
The Service surfaces what frontier AI models have been observed to
say in response to a fixed corpus of finance-related questions. Those
statements include:
Numerical "Prophecy Scores" derived from search persistence, AI consensus, training-data footprint, and short-term price momentum (see /methodology).
"Drift" measurements describing how those rankings shift across model versions.
"Fracture" alerts describing when models disagree.
These are descriptions of model output. They are not buy, sell, or
hold recommendations. A high Prophecy Score does not predict
that an asset will outperform. A drift event does not signal that a
trade should be made. The numbers can move because a single model
vendor changed a system prompt.
3. Methodology has limits
Training-data bias. Models tend to mention assets that were widely discussed in their training corpus. NVDA, AAPL, and BTC will appear more often than equally-promising assets that the internet hasn't talked about. The Prophecy Score is not a search for hidden gems; it is a measurement of conventional wisdom.
Cutoff lag. Each model's training data has a cutoff. A model released in late 2025 may have no awareness of a Q1 2026 earnings collapse. We surface model release dates wherever possible, but the lag is real.
Provider variability. Model behaviour can change without a version bump (e.g. system-prompt updates, safety-finetune patches). The "consensus" you see today may not be reproducible next week.
Cost-bounded coverage. A full sweep across every model is expensive. We may run partial sweeps or sub-sample the corpus. Coverage metadata is shown on each result; absence of a model from a result does not imply that model disagrees.
Asset extraction. We use structured prompting to extract tickers from model prose. The extractor occasionally mislabels (e.g. APPL typo, ambiguous proper nouns). False positives are filtered against an exchange list but cannot reach zero.
4. Markets are not measurable
Past model agreement does not guarantee future asset
performance. A pile of intelligent observers can be coordinated and
coordinated and still wrong. The Service is best understood as a
high-resolution mirror of consensus, not a predictor of price.
5. Conflicts of interest
thesignals labs, inc. does not currently take affiliate or referral revenue from brokers, exchanges, or asset issuers. If that changes, this section will be updated and existing customers will be notified.
The founders and employees may personally hold positions in any of the assets surfaced by the Service. We disclose this so you can weigh it. We do not trade on the basis of unreleased drift data ahead of customers.
The Service may, for engineering reasons, query some Model Providers more frequently than others. That weighting is a function of cost and rate limits, not of editorial favouritism.
6. The "Singularity Portfolio"
The "Singularity Portfolio" view in the application is a
research illustration showing how a portfolio would have
looked if it tracked the Demand surface verbatim. It is not a managed
fund, not a robo-advisor, not a discretionary account, and not
available for investment. We do not transmit any holding to any
broker. We do not execute trades. Use of this view does not create
an advisory relationship between you and us.
7. Push alerts and "Trend Fracture" notifications
Push alerts are informational. Receiving an alert is not a signal to
trade. We send alerts within best-effort timing; we do not guarantee
delivery within any specific window, and we are not liable for
decisions made on the basis of alert latency.
8. Third-party model providers
The Service queries large language models operated by OpenAI,
Anthropic, Google, Meta, xAI, Mistral, DeepSeek, and other vendors.
Each provider's outputs are subject to that provider's own terms of
service and content policies. We do not endorse any model's output;
we surface it.
9. Geographic limitations
The Service is operated from the United States. Some content may be
restricted in jurisdictions that regulate the publication of asset
rankings or investment commentary. You are responsible for
determining whether your use of the Service is lawful where you
reside.
10. Not professional advice
Nothing on this Service is legal advice, tax advice, accounting
advice, or psychological advice. For any decision with real-world
consequences, consult a licensed professional.
11. Changes
We may update these disclosures as the product evolves or as we receive
counsel guidance. Material changes are announced in-app and by email.