Reading Seven Technologies Through Both Models

TheThe Hype Cycle as a Second-Order Cybernetic System proposes three shapes a technology's publicity-versus-utility curve can take — recovers, recycles, or dies in the Trough — and Placing a Technology on the Half-Life Scale gives a rubric for where a technology sits on the separate foundational/applied/ephemeral axis. Seven real, dated cases below apply both at once. Two of the seven never appear on the diagram at all, and that absence is itself the point.

Positions are representative placements illustrating each technology's documented trajectory, not measurements on a common numeric scale — Gartner's own Hype Cycle charts work the same way. Not shown: relational databases and transformer architecture, discussed below, neither of which ever entered a hype cycle at all.

Expert Systems: Died in the Trough, as a Product

CommercialCommercial rule-based expert systems drove much of the applied-AI boom through the early-to-mid 1980s. The category collapsed with the specialised Lisp-machine hardware market in 1987 and the failure of expert systems to scale past narrow domains — a central trigger of the second AI winter, alongside the cancellation of DARPA's Strategic Computing Initiative and Japan's Fifth Generation Computer Project, as Does a Research Field Sleep? covers in full. The distinction worth holding onto: the field of AI recovered, through entirely different technical lineages (backpropagation, support-vector machines, eventually deep learning) that were quietly maturing through the same winter. Expert systems, specifically, as a commercial product category, did not — almost nobody builds a classical 1980s-style expert system today. This is the applied/ephemeral-tier pattern from the classification rubric playing out exactly as predicted: the category with high substitute pressure and shallow fan-in died; the foundational work underneath it, insulated from that pressure, kept going.

3D TV: Died in the Trough, With No Field Underneath It

3D television is the cleaner version of the same death, because there was no deeper foundational field quietly surviving underneath it to complicate the story. The 2009 release of Avatar drove manufacturers to push 3D-capable sets hard from 2010 onward: 24 million 3D-capable TVs sold worldwide in 2011, rising to over 41 million in 2012, when 3D sets accounted for 23% of total TV sales revenue¹. Sales then fell every single year from that peak. Vizio stopped making 3D sets in 2013; Samsung held on until 2016; Sony and LG dropped it from their entire lineups by 2017. No major manufacturer makes a 3D TV today. The mechanism matches the classification rubric directly: 3D TV had a close, rapidly-improving substitute (4K and HDR delivered a visible quality improvement without asking the viewer to wear anything) and essentially no fan-in of its own — nothing else in the technology stack depended on it. Once a substitute with equivalent publicity arrived, there was nothing underneath 3D TV to keep any part of it alive.

VR: Recycled, Not Recovered

VirtualVirtual reality is the cleanest recycling case available, because it's this well-documented twice over. Nintendo's Virtual Boy launched in 1995 and was discontinued within a year; consumer VR generally faded from view through the early 2000s. Fenn's own first Hype Cycle diagram, published that same year, 1995, already placed VR in the Trough of Disillusionment² — a technology can enter this framework already past its Peak. A second, unrelated wave began with the Oculus Rift's 2012 Kickstarter campaign (Facebook's subsequent $2 billion acquisition of Oculus is the clearest single publicity event in this whole case-study set), reached the Slope of Enlightenment by 2017 in Gartner's own tracking, and by 2018 had been removed from Gartner's chart entirely — officially because the underlying technologies were judged mature enough to no longer count as "emerging." The two waves are not one continuous curve; the second was triggered by real capability gains (display, tracking, latency) the first wave didn't have, which is exactly what distinguishes recycling from a simple rebound off the same peak.

Blockchain: Still Mid-Cycle

EnterpriseEnterprise blockchain has a clean, well-documented single-cycle trajectory largely because Gartner tracked it explicitly, year over year. It reached Gartner's own Peak of Inflated Expectations around 2016, was assessed as having entered the Trough of Disillusionment by 2018 (with Gartner's own analysts at the time projecting a Plateau of Productivity within roughly a decade), remained there through 2019–2020 as most applications stayed stuck in "experimentation mode," and by 2021 attention had shifted toward narrower, real use cases — trade finance and cross-border payments specifically — consistent with early movement onto the Slope. This is the pattern the "recovers" category predicts: publicity crashed hard, but utility, in the narrower form the corrected model actually supports, kept accumulating underneath it.

Frontier LLMs: Unresolved, by Design

TheThe Half-Life of Knowledge already covers the specific claims in circulation about frontier AI's rate of change; the honest position here is the same one that page takes about the underlying capability curve — genuinely fast-moving, and not yet possible to classify into "recovers," "recycles," or "dies" with anything like Chow et al.'s citation-half-life rigor. What can be said: publicity is currently very high, several of the loudest claims in circulation are ephemeral-tier claims dressed in foundational-tier language (per the classification rubric's own point about interface churn), and the second-order framing above predicts that whatever correction eventually happens will be a correction to the community's current model of what this technology can do, not necessarily a reversal of the technology's own underlying trend.

Not Shown: The Technologies That Never Entered the Cycle

RelationalRelational databases³ and transformer-based attention⁴ share the profile the classification rubric predicts should sit outside the Hype Cycle altogether: high fan-in, little effective substitute pressure at the level of the core concept, and a concept that's stayed essentially stable while everything built on top of it churned constantly. Neither ever had a Gartner-style Peak, because neither ever needed the kind of marketable novelty a hype cycle runs on — both accreted into the field's working assumptions instead of launching. Their absence from the diagram above isn't a gap in the data; it's what the rubric said would happen to something foundational.

Related Topics

• The Hype Cycle as a Second-Order Cybernetic System — the three outcome shapes these seven cases are sorted into.
• Placing a Technology on the Half-Life Scale — the rubric behind the foundational/applied/ephemeral judgements made above.
• Does a Research Field Sleep? Collective Consolidation and the AI Winters — the full account of the expert-systems collapse and what kept moving underneath it.
• The Half-Life of Knowledge — the frontier-LLM claims referenced above, examined on their own terms.
• Managing the Trough: For Builders and For Learners — what to actually do with any of this.

References