AI Lab — hyped AI topics, honestly graded

The hyped AI/quant topics people search for, mapped honestly to what Quantustik actually ships — grounded in a real artifact, or clearly labelled educational. Nothing here is a performance promise.

Start here

Topic 1: Quantum-mechanics-based stock forecasting (core)

Quantustik's core forecasting engine — a Schrödinger-equation + Feynman path-integral model with publicly calibrated confidence bands, including the failures. See the full page.

Topic 2: MCP server for AI investing agents (core)

A live, hosted Model Context Protocol server exposing S&P 500 signals, forecasts, and market-conditions data so AI agents can query Quantustik directly. See the full page.

Topic 3: LLM stock analysis & AI sentiment (core)

Two real LLM features layered on top of the quantitative model — a per-ticker AI summary and a news-sentiment score — framed honestly as context, not a trading signal. See the full page.

Forecasting literacy & honesty

Topic 10: Can AI predict stock prices? (educational)

The honest answer: a single point prediction ("$X by Friday") is unfalsifiable marketing — a published, testable interval forecast is the real product. See the full page.

Topic 11: What a 90% confidence interval actually promises (core)

A 90% confidence interval is a coverage promise, not a difficulty rating — about 90% of realized prices should land inside the band over time, and it says nothing about direction. See the full page.

Topic 12: Calibration vs. directional accuracy: why we publish one and not the other (core)

A well-calibrated confidence band and a profitable trade direction are two different questions — we publish coverage continuously and gate any win-rate behind a pre-registered track record. See the full page.

Topic 13: Where our model fails: structural breaks (core)

Named failure, not a hidden one: our confidence bands for AVGO and BRK-B at the 1-year horizon cover far fewer realized outcomes than the aggregate figure suggests — both went through a structural break the model cannot see coming. See the full page.

SEC filings intelligence

Topic 14: SEC Form 4 insider trading signals (core)

Corporate officers, directors, and large owners must report their own trades on SEC Form 4 within days of the transaction — a real paper trail, but a laggy and incomplete one on its own. See the full page.

Topic 15: 13F whale tracking (core)

Form 13F tells you what large institutional managers held at quarter end — weeks after the fact, and only the long equity side of their book. See the full page.

Topic 16: Schedule 13D/13G activist alerts (core)

Crossing the ownership-disclosure threshold triggers a Schedule 13D or 13G filing — which one tells you whether the buyer wants to influence the company or just hold the stock. See the full page.

Market structure

Topic 17: Market Conditions detection (experiment)

A market-conditions classifier tries to label the current backdrop as risk-on, risk-off, or somewhere in between — a genuinely useful framing, and one that is inherently uncertain and slow to catch turning points. See the full page.

Topic 18: S&P 500 sector rotation heatmap (experiment)

A live view grouping the S&P 500 by GICS sector — realised returns over selectable windows plus average expected growth — portfolio context, not a timing signal. See the full page.

Topic 19: S&P 500 correlation map — risk lens (experiment)

A live correlation heatmap framed as a concentration-risk lens — a metaphor for co-movement, not a prediction of it, and not literal quantum entanglement. See the full page.

Trade execution & risk

Topic 20: Position sizing with confidence bands (core)

How a calibrated forecast band's width becomes a stop distance, a reward-to-risk ratio, and a bounded position size — process, not promise. See the full page.

Topic 21: Take-profit ladders and volatility-proportional trailing stops (core)

A live TP1/TP2/TP3 exit ladder and a volatility-scaled stop, with a break-even stop advance after the first target — untested for P&L publicly. See the full page.

Topic 22: Anatomy of a trade plan (core)

A real trade plan needs an entry trigger, an invalidation level, a take-profit ladder, and a risk/reward gate — and a refusal to hand out an entry when the setup does not clear the bar. See the full page.

Developers & agents

Topic 23: Stock market MCP server (core)

A real, hosted Model Context Protocol server for stock market data — streamable-HTTP, no API key required, a public discovery manifest, and a growing tool set for signals, forecasts, and trade plans. See the full page.

Topic 24: Connect Claude to live stock data (core)

A step-by-step tutorial for pointing Claude (or any MCP client) at Quantustik's live stock market MCP server: get a key, add the server config, and run a guided risk-check prompt. See the full page.

Topic 25: S&P 500 signals API for developers (core)

A public v1 REST API for S&P 500 signals and forecasts, documented by a live OpenAPI schema — keyless for light use, metered for higher throughput, read-only by design. See the full page.

Educational research only — not investment advice.