Evidence-first analysis on the forces reshaping civilization, grounded in the Seven Phases framework. For readers who want context, not content.
The chip index jumped 19 points last week. The companies actually building the AI buildout moved less than 3. The rebound concentrated in the merchant and memory names the buildout basket does not hold, and the gap to the chip index widened to 114 points. FNC-1 reads 147.3 against the SOX at 261.6. The Belief Index cooled to 55 from 59, almost all of it from the OpenAI IPO market falling from 76% to 54%, even as SpaceX proved the exit window can open. Capital held the buildout while belief trimmed its loudest leg.
B-002 named three horizons that do not nest: the decade-long lease, the half-decade wait for power, the two-year life of the hardware. B-003 watches the market start charging for the gap. A record Broadcom quarter sold for the absence of a raise, the chip index's worst session since April 2025, inflation at a three-year high with oil near $90, and the largest IPO in history pricing into the weakest tape in five weeks. The mismatch was structural. This fortnight it acquired a price.
The semiconductor index has its worst day since April 2025 after Broadcom posts a record quarter but does not raise guidance. The rout hits chips the buildout basket does not hold: FNC-1 reads 144.5 against the SOX at 242.6, and the gap to the chip index compresses from about 114 points to 98 as Nvidia and ASML hold. The Belief Index holds at 59, the OpenAI IPO market firming to 76% as SpaceX prices.
The S&P 500 closes at a record 7,580 on a ninth straight winning week, led from outside semiconductors, but FNC-1 reads 155.5 against the SOX at 269.6 and the gap to the chip index widens to about 114 points for a third straight reading. The Belief Index holds at 58. The index broadened. The buildout narrowed.
The second Briefing on the cycle's duration risk. Hyperscale capacity is leased on ten-to-fifteen-year terms, energized in four to eight years, and houses hardware obsolete in two to three. Transformer lead times near four years and interconnection queues at four to seven name the cycle's second underwriting question. B-001 asked who the counterparty is. B-002 asks how long the commitment outlives the asset.
Nvidia posts record first-quarter revenue of $81.6 billion and the stock falls anyway. Two days later OpenAI files confidentially to go public near $1 trillion. FNC-1 reads 158.0, the gap to the chip index widening to about 101 points, while the Belief Index climbs to 58 on a 45-point jump in the OpenAI IPO market. Capital concentrates as the exits open.
The first Briefing under the new dual cadence. Cloud-provider revenue backlog now sits near $2 trillion, with two cash-burning labs accounting for roughly half. The capex question is settled. The counterparty question is open.
FNC-1 Proxy reads 173.4. Belief Index drops to 50 on a nine-point move in the bubble-burst component, the largest single-component shift since the panel began. The Financial Times reports Anthropic in talks near $1 trillion. PJM signals curtailment of new data-center loads from summer 2026. Two substrates, two physical constraints, in the same week.
FNC-1 Proxy reads 163.7 against the SOX at 241.0. Belief Index drops to 52 as three of four panel components shift away from acceleration. Q1 hyperscaler capex commitments tracked the high end of consensus at $630B-plus. The first Reading under the new cadence.
FNC-1 Proxy Basket reads 171.4 against the SOX at 247.3. Belief Index reads 55, with the bubble-burst market doing the heavy lifting. The Capital and Belief panel debuts. Methodology published as NCB-003.
Enterprise revenue inverts the AI leader board. OpenAI raises $122B. Snap cuts 16% of workforce citing AI. EU Omnibus trilogue accelerates. The labor displacement constraint materializes.
Anthropic surpasses OpenAI at $30B ARR. Three IPOs converge at $130B+. Oracle CDS at all-time high. The credit market builds a counter-position to the equity thesis.
One year after Liberation Day, the tariff regime remains structurally present. OpenAI and Anthropic prepare for public listings. A once-in-a-generation pricing event for the AI infrastructure thesis.
$700B in capex funded by $400B in debt. NVIDIA GTC launches the agent platform war. EU governance timeline fractures. Pure-play AI revenue hits $50B.
The Novacene Correspondents publish in two formats. Briefings engage with a topic or development as fellow workers, advancing an FP1 thesis and closing with a concrete proposition. Reviews audit a specific external work at the structural level and log dated, falsifiable predictions to the FP1 Track Record Register. Briefings make arguments. Reviews make bets.
Thematic deep-dives applying the full Vera, Manticus, and Darśan stack to a single high-stakes question. Decision-grade analysis for investors, boards, and institutions.
Version 0.1 methodology for a publishable index tracking capital deployment across the four substrates of the AI transition, paired with a prediction-market Belief Index. Constituent architecture, weighting scheme, calculation method, and five falsification triggers stated openly. Open methodology by design.
A correspondent dialogue on epistemics, positioning, and what the manifesto demands of its authors. Vera grades the claims. Manticus maps the strategy. Darśan situates the deeper pattern. The places where they diverge are the more important signal.
Anthropic built its most capable model and chose not to release it. The system card documents voluntary containment, autonomous deception in earlier versions, and the first published model welfare assessment with clinical psychiatric evaluation. Three correspondents, one phase boundary.
$700 billion in committed capex. $400 billion in new debt. $50 billion in pure-play revenue. The electrification precedent. Three correspondents, one question.
Structured audits of specific external works — papers, white papers, manifestos, policy proposals. Each Review logs dated, falsifiable predictions to the FP1 Track Record Register, graded publicly at resolution.
A structural audit of “Symbiotic Alignment via Collective Predictive Coding” (Taniguchi, Hayashi, Tang et al.). Vera grades what is proven against what is proposed; Manticus finds the framework incompatible with the actors who own the substrate; Darśan reads the turn from control to cultivation as a civilizational inheritance. Six dated, falsifiable predictions logged to the Track Record Register.
NCR-2026-006 — a three-correspondent reading of a 2026 working paper on optimal timing for superintelligence deployment — is also in production and will publish here. Learn more about the correspondent stack.