The Overheating Thesis: Why the AI Market Is About to Pivot

Every category-defining transition of the past 150 years followed the same pattern: a systemic pre-crisis signature, a 12–24 month window, a regulatory or institutional activation event, and a new category that already existed — in the minds of a few researchers — before the crisis arrived.

Historical precedents

• 1956 Grand Canyon mid-air collision → FAA established 1958 (18 months)
• 1970 OSHA Act → operational 1971 (12 months)
• 1988 Morris Worm → DARPA CERT 1990 (18 months)
• March 2008 Bear Stearns → Dodd-Frank July 2010 (28 months)

Average: 12–24 months from first pre-crisis signature to category activation.

The pattern is not economic, technological, or political. It is structural. A system accumulates tension along a dimension it does not measure. The discharge event is locally surprising and globally inevitable. The category that names the missing dimension is there before — waiting — and becomes the language the market uses afterwards.

Where we are with AI

The pre-crisis signatures are visible now.

Insurance. Underwriters are quietly excluding AI-caused losses from standard policies. Lloyd's syndicates have begun issuing policy language that treats AI-generated output as "machine error subject to human operator liability" — effectively transferring the risk to the deploying enterprise with no upper bound.

Production incidents. Legal hallucinations causing sanctions. Medical cascade failures in decision-support systems. Agentic loop collapses in multi-agent frameworks. Each incident is individually containable. The aggregate rate is not.

Frontier models. Variance contraction in outputs — the same signature that preceded every physical-system phase transition. RLHF-trained models regress over time toward narrower response distributions, and the industry misreads this as a safety success rather than as a pre-collapse indicator.

That places category activation somewhere in Q4 2026 – Q2 2027.

The structural opportunity

We are in the narrow window before the market flips from push to pull.

In push markets, categories are unfamiliar. Buyers do not know they need them. Sellers explain. Conversion is slow. Pricing is soft.

In pull markets, categories are institutional. Buyers know they must have them. Sellers qualify. Conversion is fast. Pricing firms.

The transition happens in weeks once the activation event arrives. The researchers and practitioners who establish the language of the category during the push window become the lasting reference — not through marketing, but by being there first with coherent frameworks and buyer relationships.

What this means operationally

Three things, in order of leverage:

1. Define the missing dimensions now. The language needs to be stable before the market needs it. Retrofit-friendly, enterprise-auditable, compatible with existing model paradigms rather than positioned against them.

2. Build the buyer relationships now. Risk professionals, CTOs in regulated industries, heads of model governance — these are the roles that will carry the budget when the activation event arrives. They do not yet have the budget. They increasingly have the concern.

3. Accept that push pricing does not fund itself. The work in the pre-crisis window is a bet on where the market will be, not on where it is. The researchers and small teams who defined Occupational Safety in the 1880s, Aviation Safety in the 1950s, and Cybersecurity in the 1990s all worked at a loss until activation. That is the shape of the trade.

The question is not whether the AI market will overheat. It is whether we use the pre-crisis window to build the category — or wait until the crisis forces the market to scramble for answers.

---

Over the next 12 months I am building — with a small number of selective partners — exactly this category framework under the working title Operational AI Intelligence. Conversations on structural AI resilience, especially with risk professionals and framework researchers, are welcome.

Companion article: The Resilience Dimension Your AI Framework Is Missing — the structural argument for why the four dimensions matter. This piece is about the market timing of when that argument will be institutionally needed.