Frequently Asked Questions

Bubble Trouble is an educational monitoring tool that tracks six signals designed to detect whether the AI infrastructure investment cycle has entered bubble territory. It uses a two-horizon architecture — structural indicators to detect if a bubble has formed, and peak detection indicators to confirm if the unwind has begun. It produces an alert level from 0 (All Clear) to 3 (Cascade Confirmed).

An independent researcher interested in market structure, capital cycles, and the history of technology investment bubbles. The project is not affiliated with any financial institution, hedge fund, or investment advisory firm.

Absolutely not. Bubble Trouble is an educational and analytical project. It does not recommend specific trades, portfolio allocations, or investment strategies. The action recommendations tied to alert levels are theoretical illustrations, not personal advice. Always consult a licensed financial advisor before making investment decisions. See the full disclaimer.

Different signals update on different schedules. Market-based signals (semiconductor leadership, market breadth, HY spreads) are assessed weekly. Capex vs. revenue divergence is updated quarterly following earnings reports. Margin debt data is monthly with approximately a 6-week reporting lag from FINRA. Private credit stress indicators are assessed weekly based on public information.

Several scenarios could cause the system to produce incorrect signals:

• The AI capex cycle is truly justified. If AI revenue growth eventually catches up to or exceeds capex growth, the structural divergence signal would remain Green despite massive spending — because the spending would be rational.
• The bubble pops through an unmonitored channel. The system monitors six specific signals. If the cycle ends due to a geopolitical event, a regulatory crackdown, or a sudden technology shift, the system might not detect it until the peak signals react.
• Timing is off. The system might trigger too early (false positive) or too late (missing the peak). The persistence requirements reduce premature triggering but can delay detection.
• Private credit opacity. Signal 2 relies on publicly available information about private credit markets, which are inherently less transparent. Stress could build in ways not visible to outside observers.

MVP design philosophy. The goal is to have a small set of well-understood, non-redundant signals rather than a large set of noisy, overlapping ones. Six signals across three layers provides coverage of the key dimensions (structural overinvestment, financing stress, market confirmation) without the complexity of managing 20+ indicators where signal noise becomes a problem. If a signal can't clearly articulate what it measures and why it matters, it doesn't belong in the system.

Interpretability. When the system says "Level 2: High Alert," you need to know why — which signals are firing, which thresholds have been crossed, which special rules are active. A black-box ML model could potentially detect patterns in the data, but it couldn't explain its reasoning, and with n=3 historical bubble events, there isn't enough training data to build a reliable ML model anyway. Transparent, rule-based logic is the right choice for this problem.

Two main risks:

• False positives — the system triggers an alert, you reduce exposure, and the market continues to rise. This means you miss upside. The persistence requirements and escalation blockers are designed to reduce this risk, but they can't eliminate it.
• Unmonitored channels — the bubble ends through a mechanism not covered by the six signals. For example, a sudden shift in government AI regulation, an unexpected technological breakthrough that renders current infrastructure obsolete, or a geopolitical conflict that disrupts supply chains.

Limited. The backtest covers three events (2000, 2008, 2022), which is statistically insufficient for any meaningful confidence. The framework was designed after studying these events, introducing hindsight bias. Data gaps exist for private credit (pre-2015) and some ETFs (pre-2000). The backtest provides directional intuition — a sense that the signal types are measuring the right things — but should not be treated as evidence that the system "works." See the full backtest analysis for detailed limitations.

This is educational content. It is not designed, tested, or validated for use as a trading system. The alert levels and action recommendations are theoretical illustrations of how one might respond to different risk environments — they are not personal investment advice. If you're making investment decisions, please consult a licensed financial advisor who understands your specific situation, goals, and risk tolerance.