Methodology — Open Construction, Public Falsification

How the layers relate

Two methodology surfaces, two different jobs.

FP1 maintains two distinct methodology layers. They reference each other but answer different questions, and should be read as complementary rather than as substitutes.

How we reason

The science of the analytical method

Vera, Manticus, and Darśan are not personas. Each operates under a named scientific framework: the Free Energy Principle and Bayesian inference for Vera, Markov blankets and active inference for Manticus, phase-transition theory and morphogenesis for Darśan. Their architecture pages document the formalism behind each desk.

What we measure

The science of the instrument

FNC-1 (the Novacene Composite) and the Belief Index are research instruments, not commentary. Their methodologies cover constituent selection, weighting, normalization, rebalancing, and the falsification criteria under which they are revised or retired.

In practice the layers cross. Vera grades evidence about index movements. Manticus reads the Markov blanket of capital and belief. The instruments measure the world; the correspondents reason about what the measurements mean.

The Radar: Signal Scoring and Historical Thresholds.

NCB-004

Methodology v0.2 · July 2026 · David Lovejoy & John H. Clippinger

How an FP1 signal is constructed, scored, and graded. The paper documents the anatomy of a call, the derivation of a historical threshold from a named reference class, the Mass–Velocity–Proximity scoring model and the structure it borrows from active inference, the grading discipline across sweeps, and the conditions under which the Radar itself is revised or retired. It is the method of record behind Scenario Assessment SA-001.

Read NCB-004 →

FNC-1 architecture

Four substrates. Twenty-eight constituents. Two weighting schemes.

No single layer dominates. The thesis is that a genuine Novacene transition requires coordinated movement across compute, energy, frontier intelligence, and biological convergence. Divergence between layers is itself signal.

Layer	Representative constituents	Weight
Compute	NVDA, TSM, ASML, AMD, AVGO, MU, AMAT, LRCX, ARM	30%
Energy	CEG, VST, GEV, ETN, OKLO, SMR, CCJ	20%
Frontier	MSFT, GOOGL, META, AMZN, ORCL, PLTR	35%
Biological	RXRX, SDGR, TWST, DNA, ABCL	15%

FNC-1 is equal-weighted within layer with a 15% per-name cap. FNC-1M uses modified market-cap within layer with the same cap. Both publish in parallel, so readers can see how much of any move is megacap concentration versus genuine breadth. Quarterly rebalancing, annual constituent review.

Public falsification

The conditions under which the index is revised or retired.

Falsification criteria are stated up front, so the reader knows what would force a revision. If none trigger, the framework is doing work. If any triggers, the methodology is restructured in public.

Redundancy

The four-layer structure contributes nothing and the index is just a SOX restatement. It should be retired or radically restructured.

Decoupling

Energy decouples from Compute for 18+ months. Either the energy thesis is wrong, or compute is scaling without the physical constraints assumed. The layer architecture is revised.

Thin signal

Biological shows no signal for 24+ months. The convergence thesis is structurally wrong, or the public-market instruments are too thin to carry it. The layer is retired rather than pretended to work.

Belief overlay failure

The prediction-market overlay shows zero predictive relationship with FNC-1 returns at any lag up to 12 months. The juxtaposition becomes aesthetic rather than analytical, and is dropped.

Structural shift

Major private frontier labs go public (OpenAI, Anthropic, xAI). The index must be restructured to incorporate them; continuing without them would be negligent. New constituents enter at the next quarterly rebalance.

Stated limitations

What the instrument cannot see.

Every measurement instrument has a blind spot. FP1 names them rather than hides them.

Public market bias

The most important Novacene actors are private. The index necessarily understates the transition until major frontier labs go public, at which point the methodology is restructured (Trigger 05).

US / Taiwan concentration

Chinese AI and compute are structurally excluded by sanctions, ADR delistings, and data-reliability issues. A real blind spot, not a methodological choice we endorse.

Reflexivity, reframed

If the index becomes cited, capital flows toward its constituents. The standard finance critique calls this distortion. FP1 treats it as empirical: the test is whether the system becomes better calibrated to first principles, not whether the instrument stays passive. See NCB-003 § IX.

Selection judgment

Constituent selection is judgment-based. The annual review is documented in public changelogs. No claim of mechanical objectivity.

Methodology track

Published, in development, queued.

Each instrument that joins the FP1 stack is documented here with the same conventions: open construction, named falsification criteria, stated limitations.

NCB-004 · Radar Methodology Paper

Signal anatomy, historical-threshold derivation (reference class, value, falsifier, freeze), Mass–Velocity–Proximity scoring, grading discipline, and falsification of the method itself. Authors: David Lovejoy & John H. Clippinger.

Published

NCB-003 · FNC-1 Methodology Paper

Constituent architecture, weighting schemes, rebalancing logic, falsification framework, prediction-market overlay design. Authors: David Lovejoy & John H. Clippinger.

Published

FNC-1 Live Index

Daily-calculated index series with backtested history. Hosted chart, public methodology version log, change log of constituent revisions.

In development

Belief Index Methodology

Standing market panel, normalization rules, weighting across venues (Kalshi, Polymarket, Manifold), and the divergence calculation against FNC-1.

Queued

Substrate Sub-Index Decomposition

Independent reads on Compute, Energy, Frontier, and Biological layers, and the cross-layer correlation tracker that triggers the decoupling criterion.

Queued

Annual Methodology Review

Each January, FP1 publishes a review of every active instrument: what triggered, what changed, what was retired. The audit lives publicly alongside the methodology.

Annual

Open questions

What we are still working out.

Methodology is iterative. These are questions FP1 is actively working through, named openly to invite scrutiny from institutional readers, methodologists, and other thought partners.

Is the correlation test against the semiconductor index the right null hypothesis, or is there a stronger one? Equal-weighted S&P 500 is one alternative; the MSCI ACWI IMI Future of AI Index is another.

Should the Belief Index publish as a single normalized acceleration score, or as a panel by question category (capability milestones, governance events, AGI timing)? A single number is cleaner. A panel is more honest.

When private frontier labs go public, does the methodology accommodate them at the next quarterly rebalance, or is an extraordinary revision warranted? The methodology currently allows extraordinary revision on material regime shifts. The trigger language may need tightening.

Should the substrate architecture eventually be reorganized around first principles (free-energy minimization, autopoietic alignment, least action) rather than the current Anthropocene-categorical layers? The current view: substrates are load-bearing scaffolding for measurability. The first-principles lens belongs to the correspondent stack, where it informs reading without constraining the instrument.