Sentinel Alpha is an AI market intelligence engine built around one question: why are markets moving? Price tells you what happened. Sentinel Alpha works on the layer underneath — the narratives forming in real time, the psychology of the crowd trading them, and the places where the story and the price action have quietly come apart.
The four layers
Four times each trading day, the engine sweeps social signals, news flow, and market internals, distilling them into discrete units of evidence.
A reasoning model clusters that evidence into the narratives actually moving markets, ranks them, tracks each one through its lifecycle, and detects the earliest "seeds" of stories that haven't become headlines yet.
Every pre-market, the deepest analysis layer studies what the crowd believes, what it may be missing, and where positioning has diverged from evidence — then writes the daily synopsis and issues predictions across three time horizons, refreshed once daily.
After the close, a grader scores past predictions against actual market outcomes and publishes the result. Systematic errors feed back into the engine's calibration.
The self-grading philosophy
Market commentary has a memory problem: the hits get amplified and the misses evaporate. Sentinel Alpha is built to make that impossible for itself. Every prediction it publishes carries an ID, a timeframe, and a confidence level — and when that timeframe elapses, the grader scores it against what the market actually did and publishes the grade.
Self-grading isn't just accountability theater. Calibration data flows back into the engine: if it's consistently overconfident in a particular kind of call, its future confidence in that category is dampened automatically. The system is designed to improve because it keeps score, not despite it.
The sector intelligence layer
Alongside the daily market-wide analysis, Sentinel Alpha maintains a deep weekly assessment of 14 sectors — from semiconductors and energy to crypto and frontier AI. Two sectors refresh every day on rotation, so the full map is never more than a week old. Each report asks the same question at sector granularity: what does the development evidence show, what is the market pricing, and where do the two disconnect? Those disconnects are the alpha section — and the strongest ones are nominated back into the main engine as candidate seeds.
How to read the Pulse dashboard
The Pulse dashboard is where everything the four layers produce lands, top to bottom in reading order. Here's what each section means and how to use it. (Tap the small ⓘ beside any section title on the dashboard for a quick reminder.)
Morning Read
Sentinel Alpha's flagship analytical output — a detailed written analysis produced once daily by the system's deepest reasoning layer. It goes beyond what the data says on the surface to examine what the crowd believes, what the crowd may be missing, and where the asymmetric opportunities lie. The dateline shows exactly when it was written and what the market had done by then.
Overall Sentiment
The sentiment spectrum reflects Sentinel Alpha's current read on market psychology, scored from -1.0 (extreme fear) to +1.0 (extreme greed). This isn't just a price-based indicator — it synthesizes narrative tone, crowd positioning, and institutional signals into a single score.
The assessment text explains the key factors driving the current score. Consensus strength measures how unified market participants are — high consensus (above 70%) can itself be a contrarian warning signal.
Active Narratives
These are the top 5 market-moving stories right now, ranked by strength and relevance. Each narrative represents a cluster of related signals — news reports, social posts, analyst commentary, and data releases — that Sentinel Alpha has identified as driving market behavior. The takeaway sentence and crowd-positioning bar are visible on every card; open Read Full Analysis for the complete reasoning. Each narrative is assigned a temporal tier:
- Tier 3 (Short-Term) — Playing out in hours to days: earnings reactions, data releases, breaking news
- Tier 2 (Medium-Term) — Developing over days to weeks: policy shifts, sector rotations, evolving geopolitical situations
- Tier 1 (Long-Term) — Structural themes over weeks to months: regime changes, macro cycles, institutional repositioning
Predicted Moves
Sentinel Alpha generates predicted market moves across all three temporal tiers. Each prediction includes a direction (Bullish, Bearish, Neutral, or Volatile), estimated magnitude per index, timeframe, confidence level, the key catalyst driving the call, and — where stated — the condition that would invalidate it. The scenario bar shows best, base, and worst cases sized by probability; tap a segment for the detail.
Predictions and contrarian alerts are generated once per day during the deep analysis and carried forward through the day — the timestamp on each section shows exactly when they were generated. Narratives, sentiment, and seeds refresh with every scan cycle.
Contrarian Alerts
Early warning signals that fire when Sentinel Alpha detects conditions historically associated with market reversals or surprises. This section is intentionally hidden when no alerts are active — if you see it, pay attention. There are four alert types:
- Narrative-Action Divergence — The dominant story says one thing, but price action is doing the opposite
- Consensus Crowding — Too many participants are positioned the same way, creating reversal risk
- Institutional-Retail Divergence — Smart money and retail are moving in opposite directions
- Seed Consensus Saturation — An emerging narrative has become so widely accepted it may already be priced in
Each alert carries a severity level (low, medium, high, critical) and lists the specific tickers most exposed.
Seed Tracker
Seeds are the earliest detectable signals of emerging market narratives — ideas, themes, or catalysts that haven't yet reached mainstream consensus. Think of them as narrative intelligence gathered before it becomes a headline.
Conviction (1–10) measures how strongly the evidence supports a seed. Velocity tracks whether conviction is rising or falling. Seeds in the Consensus stage are a warning — narratives that everyone agrees on often precede reversals. Seeds that go multiple consecutive scan cycles without new evidence are automatically killed.
Sector Intelligence
A weekly deep assessment of 14 sectors, refreshed on a rotation of 2 per day (weekends included). Each report gives a sector-level sentiment score (−5 to +5), the key developments, what the market is currently pricing, and — the alpha section — the disconnects where evidence and price action have come apart. Every entry shows when it was last refreshed, and the full report shows when it's next due.
Performance Dashboard
Sentinel Alpha grades its own predictions against actual market outcomes. The dashboard shows the trailing 30 days. Two things are actually graded: directional accuracy (did the market move the way the prediction said?) and narrative accuracy (was the stated reasoning right?). From those grades the engine computes a calibration score — whether its confidence levels are honest. The remaining tiles, seeds confirmed and average foresight, track how early the engine spotted narratives that later proved out; they're statistics the engine keeps, not grades it assigns.
The calibration chart breaks this down by confidence bucket. Bars above the target line suggest the system is underconfident; bars below suggest overconfidence. Performance data builds over time — expect this section to be sparse in the system's first days of operation.
Key Concepts
Atoms are the smallest unit of market information in the system — a single signal captured from social media, news, or market data. Atoms are the raw material that gets processed into narratives and seeds. On the dashboard, each narrative's sources badge counts the distinct atoms behind it.
Temporal Tiers classify every narrative, seed, and prediction by time horizon. Tier 3 is hours-to-days, Tier 2 is days-to-weeks, Tier 1 is weeks-to-months. This separates daily noise from structural market shifts.
Self-Grading Loop — Sentinel Alpha doesn't just make predictions, it grades them against actual outcomes and calibrates itself over time. If the system is consistently overconfident in one area, it automatically dampens its confidence. This creates an engine that improves with every trading day.
Who it's for
Sentinel Alpha is for people who want context, not signals: traders and investors who already do their own work and want a disciplined, tireless read on the narrative layer of markets — one that timestamps everything, states its confidence, and owns its misses in public.
It is not financial advice, and it isn't trying to be. The engine is experimental, its output is generated by large language models, and its predictions can be wrong — the scoreboard exists precisely so you can see how often. Read the full disclaimer before using it in any decision.