Appears above the ticker grid once all stocks have loaded. Shows the three stocks with the strongest positive composite signals, ranked by risk-adjusted score. Each card shows the ticker, live price, composite score, and a two-sentence rationale generated by Claude explaining specifically why that stock is ranking today and what the key risk is. Tap any card to open the full analysis. Updates on every refresh.

The grid shows 26 AI-related stocks across six categories. Each card shows the live price, today's % change, and a coloured bar — green means a positive signal, red means negative. Tap any card to load the full analysis below. Stocks load in batches over about 4 seconds on first open.

Filter the grid by sector:

Chip Design — fabless chip designers: NVDA, AMD, AVGO, QCOM
Semicon Mfg — chip manufacturers and equipment: TSM, ASML, AMAT, MU
Cloud Platform — hyperscalers and cloud AI: MSFT, GOOGL, AMZN, META, ORCL
AI Software — enterprise AI software: PLTR, NOW, CRM, SNOW, AI
DC & Energy — data centre infrastructure and power: CRWV, VRT, ETN, VST
International — Chinese AI names: BABA, BIDU

Up to three alert banners appear at the top of the Signals tab:

Red — Hormuz Active: The Strait of Hormuz closure is hardcoded as an active crisis since March 2026. Shows the causal chain impact on the selected stock.

Amber/Red — Earnings: Appears when earnings are within 14 days. Shows the date, EPS estimate, and a confidence penalty — signals are less reliable before earnings.

Colour-coded — War Risk: Appears when the war risk score for the selected stock is elevated. Shows the risk level and net impact.

Below the stock header, Claude AI reads the last 7 days of news headlines and scores overall sentiment from -1 (very negative) to +1 (very positive). The coloured tags show detected themes like "Earnings Beat" or "Regulatory Risk". If the Anthropic API is blocked on your network, sentiment defaults to neutral.

Controls how much the geopolitical risk score influences the composite signal. At 0% geo is ignored. At 50% it dominates. The default is 20%. Increase during periods of high geopolitical tension — the Hormuz and Taiwan situations in particular warrant a higher geo weight for infrastructure stocks.

At the top of the Signals tab — probability estimates for upward price movement over 3, 5, and 10 days, with a target price for each.

How to read it: 65% means conditions currently resemble past situations where the stock moved up 65% of the time. Not a guarantee.

Confidence shows how many of eight signals agree. HIGH = most signals aligned. LOW = signals mixed.

Earnings penalty: Probabilities are automatically reduced when earnings are within 14 days — signals are structurally less reliable in that window.

Probabilities are capped at 78% to avoid false certainty.

Shows ten stat boxes for the selected stock:

Composite — overall signal (above 0.55 = buy threshold, below 0.25 = exit)
Signal — plain English: BULLISH → LEAN BULL → NEUTRAL → LEAN BEAR → BEARISH
Sentiment — Claude's news reading, -1 to +1
Social — Reddit/StockTwits mention volume and sentiment
Analyst — Wall Street consensus: % buy, hold, sell
Insider — Net SEC Form 4 insider buying vs selling
Geo — geopolitical impact on this stock
DC — data centre activity signal impact
War Risk — war risk exposure impact
Price Target — analyst consensus target vs current price

The price chart shows 90 days with MA5 and MA20 moving averages. The composite chart shows when signals crossed buy/exit thresholds.

Breaks down geo risk across five categories, each showing active risk level and the stock's structural exposure:

Export Controls — US chip export restrictions to China
Taiwan Strait — TSMC supply chain risk from military tension
EU Regulation — AI Act enforcement and data regulation fines
China Revenue — risk of losing Chinese market access
US AI Policy — executive orders, congressional action, defence spending

High exposure + low active risk = fragile position. Low exposure = mostly immune.

Shows data centre and energy infrastructure signals across five categories:

Power Demand — power purchase agreements, gigawatt commitments, grid capacity
Hyperscaler Capex — data centre investment and expansion announcements
Cooling & Water — liquid cooling, water rights, thermal management
Colocation Leasing — data centre lease signings and campus builds
Energy Markets — electricity prices, renewable PPAs, transformer shortages

CRWV, VRT, and ETN have the highest DC exposure. PLTR and software names have very low DC sensitivity.

Models war and conflict risk across five categories: Broad Conflict, Supply Chain Disruption, Defence Spending, Energy Security, and Safe Haven Flows. Each stock has different sensitivity — PLTR is the only name where war risk can be net positive (defence AI contracts). TSM and CRWV are the most exposed negatively.

The net impact number shows whether current war conditions are a headwind or tailwind for this specific stock after all five categories are weighted.

Shows all registered named geopolitical scenarios and their current status:

● ACTIVE — scenario is currently occurring and full causal chain is propagating
◐ ELEVATED — recent signals detected but not confirmed active
○ DORMANT — no current signals

For active scenarios, the causal chain shows each downstream effect with its time lag (T+0d, T+3d, T+14d etc.) and the per-stock impact. Green nodes have fired; grey nodes are pending. The Hormuz closure is permanently ACTIVE based on confirmed live conditions since March 2026.

Shows what would have happened trading the composite signal over the last 90 days. Buys when composite crosses 0.55, sells when it drops below 0.25 or after 5 days. Green line = strategy, grey dashed = buy and hold. Alpha = outperformance.

The logic box shows current composite weights across all eight signal layers. Weights shift as the Bayesian updater learns from resolved predictions.

Shows average composite signal across all six sector categories. When one sector pulls significantly ahead of others, capital historically rotates toward it. The spread number shows how wide the gap is — above 0.4 is historically significant. The bar chart shows all 26 stocks side by side.

Every time you open a stock, AIbacus records the prediction and current signal values. After 3+ days it checks whether the price moved in the predicted direction. After 3 resolved predictions, it starts adjusting signal weights per stock — signals that have been right get more weight, signals that have been wrong get less. The learned weights panel appears in the prediction box once learning begins. Weights are stored in your browser and persist across sessions.

Clears all cached data and reloads everything fresh. Data is normally cached for 5 minutes to avoid Finnhub rate limits. The geopolitical, war risk, and DC signals are re-scored from fresh headlines on every reload.