技术革命周期诊断（佩雷斯框架 / Perez Tech-Cycle）
Overview
This skill operationalizes Carlota Perez's model from Technological Revolutions and
Financial Capital: The Dynamics of Bubbles and Golden Ages (2002). It turns the book's
historical narrative into a repeatable phase-diagnosis procedure: identify the
technological revolution in question, score observable evidence against the recurring phase
signatures, determine the dominant phase (and its runner-up), then map the phase to the
correct financial-capital vs production-capital logic and to concrete investment / strategy
implications.
The core insight to carry into every analysis: the crash is a scheduled phase of the
revolution, not a refutation of the technology. The deciding question is never "is it real?"
but "which phase are we standing in?"
When To Use
- User asks "现在的 AI / 半导体 / 新能源 / 某个产业处在技术革命的哪个阶段？"
- User asks "这是不是泡沫？" / "现在是不是转折点？" / "会不会有黄金时代？"
- User references Perez, 技术革命与金融资本, 金融资本, 泡沫, 黄金时代, 康德拉季耶夫长波, 创造性破坏.
- User wants a tech-/industry-cycle read to inform asset allocation, sector rotation, or
corporate strategy. - User wants to stress-test a "this time is different" thesis against 240 years of precedent.
Workflow
Step 1 — Identify the revolution and paradigm
Determine which technological revolution / techno-economic paradigm is under analysis.
Load references/framework.md for the canonical list of five revolutions and the definition
of a techno-economic paradigm (key low-cost input + infrastructure + organizing model).
If the subject is newer than ~1971 (e.g. AI, bio-tech, energy transition), treat it as a
candidate 6th revolution and note the analogy explicitly. Always state the paradigm's
"best-practice common sense" before diagnosing phases.
Step 2 — Load the framework reference
Read references/framework.md to anchor:
- The two-period / four-phase model (Installation: Irruption → Frenzy; Turning Point;
Deployment: Synergy → Maturity).
- Financial capital vs production capital behaviour per period.
- The state's role and the inequality reversal at the turning point.
- The three historical golden ages as calibration anchors.
Step 3 — Run the phase diagnosis (scorecard)
Apply the 10-criterion observable scorecard in references/scorecard.md. For each criterion,
record the current evidence and score how strongly it matches each phase signature (0–3).
Two ways to execute:
- Script (preferred, auditable): generate a blank matrix with
python scripts/phase_scorecard.py --template scores.csv, fill scores.csv from evidence,
then run python scripts/phase_scorecard.py --input scores.csv --name "<主题>". The script
sums per-phase columns, reports the dominant phase and its runner-up, and appends a dated
row to history.csv so the phase judgment becomes a time series.
- Manual: tally the 10 criteria by hand, keep the runner-up visible, never declare the
phase "settled" — hold the frenzy case and the synergy case side by side.
Step 4 — Map phase → implications
Use this decision rule, drawn directly from the framework:
- Irruption / early Installation: own the builders of the new infrastructure; accept
bubble volatility; financial capital leads, profits from building capacity.
- Frenzy / late Installation: watch for decoupling of paper from earnings, valuation
dispersion collapse inside the theme, issuance flood, retail euphoria — these are frenzy
signatures, not confirmations.
- Turning Point (crash): survival and balance-sheet strength dominate; the crash
transfers the installed base to stronger hands at rational prices — the precondition for
the golden age, not its negation.
- Synergy / Deployment: the winners are the users of the installed base — companies
applying the new paradigm to take margin in ordinary industries; production capital leads,
profits from using it; inequality reverses; state sets stable, credible rules.
- Maturity: paradigm exhaustion; returns migrate to finding the next "big bang".
Step 5 — Add the state & inequality lens
State which actor must intervene at the turning point (proactive state: regulation, public
investment, income redistribution). Flag whether the current setup risks a "gilded age"
(false prosperity) versus a true golden age. Note the inequality signature: top-1% share
rises through installation/frenzy and falls through deployment.
Step 6 — Emit a structured diagnosis report
Output: (1) the revolution/paradigm identified; (2) the phase read with score and runner-up;
(3) the financial-capital vs production-capital logic in play; (4) the three most decision-
relevant implications; (5) the key risk if the read is wrong (the "fighting the wrong war"
error: selling a real revolution because it crashed, or holding builder-phase exposure on
deployment-phase assumptions).
Guardrails
- The model is a heuristic, not a mechanical predictor. Each occurrence is unique;
overlaps and irregularities are normal. State this caveat in every diagnosis. - Score against evidence, not mood. The 10-criterion scorecard exists precisely to force
evidence-based scoring instead of vibes. - Always keep the runner-up phase visible — allocation should hold both the frenzy and the
synergy case until evidence resolves them.
Resources
references/framework.md— canonical framework: five revolutions, two-period/four-phase
model, financial vs production capital, state role, golden ages, theoretical lineage.references/scorecard.md— the 10 observable criteria, what evidence to collect per
criterion, and the phase signature each points to.references/examples.md— worked examples: trigger phrases, a fully filled
Frenzy-read scorecard (AI 2026), its mapped implications, a contrasting Synergy-read
scorecard (internet ~2012), a sample output report, and a no-script manual tally. Read
this first when unsure how to operate the skill.scripts/phase_scorecard.py— stdlib-only scorecard: generates a blank matrix, scores a
filled CSV, reports dominant phase + runner-up, and logs a dated history row.
