Forward 2026 scenario · modelled, not forecast
“What if the disruption I championed re-prices me?”
That is the uncomfortable case, and it deserves a straight answer. The asset that funds your life is built on the same AI thesis you helped prove — so a sector re-rating does not just hit a line item, it hits the thing you are most certain about. Conviction is not a hedge. The position can be right about the future and still be mispriced for two years.
The cost to a concentrated book
What does speed asymmetry cost a concentrated holder?
It costs you the gap between two clocks. Deployment moves exponentially; portfolio and institutional adjustment move linearly — capital cannot reallocate, lock-ups cannot expire, and conviction cannot revise at the speed the narrative shifts. The exposure is structural, not directional. This is the same speed mismatch that detonated the dot-com bust of 2000, where the technology kept winning while the equity that funded it lost 78%.
The concentration is now measurable at the index level: the ten largest S&P 500 names reached 41% of total market capitalization in January 2026, the most extreme reading in fifty years (Torsten Slok, Apollo Academy, “Extreme AI Concentration in the S&P 500,” 13 Jan 2026). If the broad index is that concentrated, a single-ticker book is a leveraged bet on one regime holding — a concentration risk we map alongside the current market environment and the parallel US debt and inflation scenario for 2026.
What does engineered resilience look like here?
It looks like a position you can still read when the narrative turns on you. You stop defending the thesis and start mapping the drawdown — the re-rating that triggers a hedge, the correlation that snaps, the carry you pay to keep optionality. This is the tail-risk options discipline applied to a single conviction holding, built on the Spitznagel approach to convexity. The conviction stays; the blind spot closes.
How it works
The Sandbox Engine
You run the scenario yourself, against your own allocation — no call, no sequence.
Model · diffusion sampling over fat-tailed paths · coverage · your concentration vs. an AI re-rating · stated 4.2% false-comfort rate
The engine compiles 50,000 macro paths and prints where the position breaks, carrying a published 4.2% false-comfort rate — the share of runs that understate realized tail loss. Kobra’s Black or White fragments one face into panels that no longer align: that is the speed asymmetry made literal — the technology recomposes faster than the portfolio holding it can.
The objection
“Isn’t this just fear-mongering about AI?”
No. The thesis may well be right; the timing and the concentration are what go un-load-tested. The diagnostic does not predict a crash — it shows you the distribution, including the runs where your tail is smaller than you feared.
3 fields · 48-hour document · no call, no sequence.
Frequently asked questions
What is the 2026 AI disruption scenario?
It is a forward scenario in which the AI sector re-rates downward even though the underlying technology keeps working. The position stays right about the future and is mispriced for one to two years while capital fails to reallocate at narrative speed. It is a stress-test input, not a forecast.
What is the structural mechanism behind it?
The mechanism is speed asymmetry: deployment compounds exponentially while portfolio and institutional adjustment move linearly. Lock-ups, mandates, and conviction cannot revise as fast as the narrative shifts, so the exposure is structural rather than directional — it does not depend on the thesis being wrong.
How likely is an AI re-rating in 2026?
No single probability is honest here; the diagnostic prints a distribution across 50,000 paths rather than a point estimate. The relevant figure is concentration: the ten largest S&P 500 names hit 41% of total market capitalization in January 2026, the most extreme reading in fifty years, which widens the tail materially.
What would trigger the scenario?
A trigger is any event that forces simultaneous repricing faster than holders can adjust — a margin disappointment in a bellwether name, a capex-return reset, or a liquidity shock that breaks the correlation holding the cluster together. The trigger does not have to be large; the concentration supplies the leverage.
How do you hedge a concentrated AI position?
You map the drawdown before you defend the thesis, then size convex protection — typically tail-risk options carrying a stated 4.2% false-comfort rate — against the specific re-rating that breaks your book. The conviction stays intact; the goal is to keep the position readable, not to exit it.