thesignals is a passive engine. We fire 4,832 financial questions — synthesized from 20 years of human search archive — at every frontier model on every release. Out come two coupled surfaces: what every model tells humans to buy, and what makes those answers structurally possible.
The corpus is fixed and read-only — 4,832 questions synthesized from 20 years of search archive. Every model release re-answers it. The answers split into two coupled surfaces. Demand pulls supply. Supply validates demand.
4,832 questions synthesized from 20 years of Google query archive. Read-only. Version-controlled. The questions don't change — only the answers do.
Answers split two ways. The demand surface ranks what humans will own (BTC, VOO, NVDA). The supply surface ranks what makes that demand possible (TSM, ASML, VST). Each links to the other.
You don't write prompts. You watch the corpus, watch the surfaces, watch the rank trajectory of every answer across model releases. The drift IS the signal.
Four signals, weighted and combined: the deeper an asset's training-data footprint, the higher its score, the more inevitable its inclusion in the consensus portfolio.
Every asset on both surfaces gets the same Prophecy Score. NVDA scores 94 because the demand surface keeps pulling it ("how do I invest in AI") and the supply surface keeps validating it ("who makes the GPUs the answers require"). The coupling is the conviction.
Built for the operator who understands that markets are now a function of model weights.
Type a question. Watch GPT-5, Claude 4.5, Gemini 2.5, Llama 4, Grok 4, Mistral Large, DeepSeek V4 and seven more answer in parallel. We extract the assets, compute consensus, and score the conviction.
Get notified the moment models start to disagree on an asset you hold. The first crack is the trade.
When GPT-5.1 ships at 9 a.m., you see what it now recommends differently — by 9:03.
Auto-rebalanced to maximum AI consensus. The assets every model recommends. The trade everyone is about to make.
Pipe consensus scores directly into your trading systems. WebSocket streams. Sub-second latency.
For twenty years, a billion humans typed the same questions into Google. "How do I get rich? What stock should I buy? Is bitcoin safe?"
That archive is now the substrate every frontier model is trained on. The questions don't change. The training data converges. The answers converge. And a billion people are about to ask AI what to buy with their savings.
They will get the same answers. We don't predict markets — we map the buying force that's already loading in the chamber.
— Cassian Vale, Founder & CEOFree trial gets you the weekly consensus. Pro gets you in front of the wave.
Weekly consensus reports across all major models. Top 10 inevitable assets. Basic prophecy scores.
Real-time consensus across every model, every update. The Singularity Portfolio. Trend Fracture alerts. API access.