Substance over hype

The belief

When a new category arrives with a new label, the first move is to name what it actually is at the layer below. If the substrate is genuinely new, the excitement is earned. If the substrate is a renamed version of something already understood, the rebrand gets refused - in writing, regardless of financial exposure to the category being evaluated.

The substrate test is a single-move procedure: strip the marketing label, name the underlying mechanism, ask whether that mechanism is new. "Blockchain" becomes "a database innovation first." "GenAI" becomes "a platform that radically reduces the cost of cognition." "Agents" become "LLMs with tools and memory." The test forces one specific question - what is actually new here - before any evaluation begins.

How to apply

1. Reduce before evaluating. When a new category claims your attention, name its substrate before forming a view. Ask: what is the mechanism underneath the label? A category that cannot be reduced to a clear substrate is noise.
2. Name the substrate in one sentence. The reduction is done when it fits in a single declarative sentence. If it requires a paragraph, the substrate has not been found yet. "A database innovation first" is the form.
3. Apply the same test inside a category as outside it. Hype operates at every level: the category claim, the product claim, the feature claim. A six-month implementation tax hidden behind "super customizable" is a failed substrate test at the product level.
4. Withhold belief until the capability claim passes an observable test. State the condition under which you would believe it: "I would catch on to the hype once Devin replaces its founders to build Devin 2.0." This is not cynicism. It is a conditional that keeps evaluation honest.
5. Refuse the rebrand regardless of exposure. The hardest application is inside a category where you have personal or financial stake. The test does not adjust for proximity. Inside-the-category honesty is the only version that counts.

What this is not

• Not "new things are never real." The substrate test is not skepticism by default. GenAI passes the test: radically reducing the cost of cognition is a genuinely new mechanism, not a rebrand. The test is neutral - it returns a verdict, not a prior.
• Not "wait for certainty before acting." Refusing the hype frame does not mean refusing to build or invest. It means acting on the actual substrate claim, not the marketing one. Agent-first runs on the claim that LLMs with tools and memory create a new serving layer - not on the label "agentic AI."
• Not applicable only to technology categories. The same test applies to PM craft, product design, and organizational claims. Anywhere a label substitutes for a mechanism, the substrate test applies.

Argues against

• "This category is different - the hype is justified this time."
• "Being skeptical of new technology is the same as being against progress."
• "The label matters for market positioning even if the underlying mechanism is familiar."

Where to go from here

If you want the root framing this belief expresses - the disposition that licenses all downstream evaluations - go to the substance-over-hype theme.

If you want to see how this belief conditions agent-first - how refusing horizontal AI hype and accepting the substrate-layer claim produced the serving-lens thesis - go to agent-first.

If you want to see the substrate test applied to product craft - spec over sprint as the anti-generative-tools verdict from a single afternoon with Google Stitch - go to spec over sprint.