Placing a Technology on the Half-Life Scale

TheThe Half-Life of Knowledge establishes three tiers — foundational, applied, ephemeral — and shows real domains sitting at each one. What it doesn't give is a way to place a new, arbitrary technology into one of those tiers before the citation-half-life data exists to measure it directly, which is exactly the situation anyone evaluating a current tool is actually in. Three structural signals do that job without waiting for years of retrospective data.

Dependency Depth and Fan-In

HowHow much else in the field is built on top of this, and how far down the stack does it sit? A foundational technology has high fan-in — many other things depend on it, directly or through several intermediate layers — and sits low in the stack, underneath the layers that change. An ephemeral technology is typically a leaf: little or nothing depends on it, because it sits at the top of the stack, closest to whatever's currently visible to an end user. Relational algebra has enormous fan-in; this month's specific ORM library wrapping it has almost none.

Substitute Pressure

HowHow many close substitutes are actively competing to displace this right now, and how fast are new ones arriving? A crowded, still-competing category all but guarantees a short half-life for any individual member of it, regardless of how good that member currently is — the pressure comes from the category's structure, not from a flaw in the specific technology. A technology with no real substitute, because it defines the category rather than competing within it, has nowhere to be displaced from.

Interface Churn Versus Concept Churn

SeparateSeparate what's actually changing: the underlying idea, or the specific surface exposing it. Gradient descent is decades old and essentially unchanged as a concept; which framework's specific method call invokes it changes constantly. When the interface churns much faster than the concept underneath it, that's a strong sign the concept itself belongs in a higher tier than the interface does — measuring only the interface's own churn rate would misclassify the whole technology as more ephemeral than it actually is.

Signal	Foundational	Applied	Ephemeral
Fan-in / stack depth	High; sits low in the stack	Moderate; mid-stack	Low or none; a leaf
Substitute pressure	Little to none; defines the category	Some; slow-moving competition	High; crowded, fast-turnover category
Interface vs. concept churn	Concept essentially static	Both move, concept slower	Interface and concept both move fast, often together

A Quick Worked Example

TheThe relational model has near-total fan-in across the entire database field, effectively no substitute for what it actually specifies (competing paradigms occupy different niches rather than displacing it), and a concept that hasn't meaningfully changed since 1970 even as every specific database product built on it has been rewritten many times over — foundational on all three signals. Transformer-based attention has, so far, shown the same pattern from a much shorter starting distance: heavy fan-in as more architectures build on it rather than around it, no real substitute currently displacing it as the dominant mechanism, and a core mechanism that's stayed conceptually stable through eight years of scaling even as every specific model built on it has been replaced multiple times. Reading Seven Technologies Through Both Models applies the same three signals across the full case-study set, alongside each one's hype-cycle position.

Why Hype Mostly Happens at the Top of the Stack

ThisThis rubric and the Hype Cycle interact in a specific way worth stating plainly: a hype cycle needs a publicity input, and publicity needs something novel and marketable enough to generate attention in the first place. Foundational-tier material rarely supplies that — a result that extends or re-scopes an existing theory accretes quietly into the literature rather than triggering a launch event, a funding round, or a magazine cover. That's why relational algebra never had a Peak of Inflated Expectations, while individual database products built on it regularly do. High-fan-in, low-substitute, concept-stable technologies mostly sit outside the Hype Cycle altogether, not because they're immune to hype but because they were never eligible for the kind of attention that starts one.

Related Topics

• The Half-Life of Knowledge — the three-tier model this page gives a classification method for.
• The Hype Cycle as a Second-Order Cybernetic System — why hype concentrates at the ephemeral end of this scale.
• Reading Seven Technologies Through Both Models — this rubric and the hype cycle applied together to real, dated cases.
• Funding Blue-Sky Research to Survive the Trough — the substitute-pressure and fan-in signals above, applied at the portfolio level to how research funding itself should be diversified.