We were a year wrong on the most important date in Chinese industrial policy.

Two corrections, both load-bearing

China's 15th Five-Year Plan (15FYP) was adopted at the National People's Congress on 12 March 2026. We had been computing every prediction's resolve-by date against the wrong NPC date — twelve months off. An audit against the National Development and Reform Commission (NDRC) announcement and the NPC archive caught the error; the fix shifted six predictions onto a tighter timeline, with the first resolution window now closing in September 2027 rather than September 2028. The same audit revisited a methodological call from the founding release. We had omitted quantum information from the forward predictions on the grounds that no firm-level demand signal exists and no industrial-fund analog has materialized. Re-read against the 13FYP-to-14FYP record, the continuity at chapter-level designation is itself a falsifiable bet. Quantum is restored as the seventh live prediction at a state-laboratory magnitude band (1–5 billion USD-equivalent).

A forecasting program that can't get its own anchor date right is unreadable

The founding release dated the 15FYP NPC adoption to 15 March 2027 in a single constant and propagated that date through six live predictions' resolve-by fields. The date is the central anchor on the entire forward register: every falsification window is computed as the adoption date plus a per-prediction lag, and every reader uses the resolve-by dates to decide when to come back and check. A twelve-month error on that anchor is not a rounding mistake. It is the wrong claim about what year the 15th plan exists in.

The error origin was simple and embarrassing: the founding dossier was drafted before the NPC session, and the placeholder date was a one-year-forward estimate that never got updated when the session actually happened and adopted the plan on 12 March 2026. The constant lived in a single line of code, was read by the page renderer, and propagated to seven public-facing date strings. Nothing in the original pipeline cross-checked the anchor against the NPC archive on each run.

Every cited 14FYP plan line resolves to gov.cn, the issuing ministry, the NPC archive, or a State Council bulletin. Press estimates appear with explicit attribution.

That guardrail covered plan lines. It did not cover the anchor date itself. This release closes the gap with a primary-source pin on the NDRC announcement of NPC adoption (12 March 2026) and adds the anchor date to the punch list of fields the pipeline cross-checks on every run.

Quantum is a continuation cascade at chapter-level designation

The founding release excluded quantum information from the forward 15FYP prediction list on a one-line justification: no firm-level demand signal, no industrial-fund analog. The audit pass reopened the question on a different framing.

The reframing is the continuity question. Quantum information was designated a frontier field in Chapter 9 of the 13th Five-Year Plan and again in Chapter 9 of the 14th. The National Laboratory for Quantum Information Sciences in Hefei was funded across both plan periods; quantum-network testbeds expanded to more than 4,600 kilometers of trunk fiber by 2024; the Jiuzhang photonic processor (2021) and the Zuchongzhi superconducting line (2023) shipped through the Chinese Academy of Sciences, the University of Science and Technology of China, and the Ministry of Science and Technology (MOST). The pattern is not that the cascade failed to compile; it is that the cascade compiles at a different magnitude band than the industrial-fund cascades, because the spending vehicle is state laboratories rather than industrial funds.

Under that framing, the relevant 15FYP prediction is whether the chapter-level designation recurs. If the final 15FYP Chapter 9 text retains quantum information at the frontier-field tier alongside the seven 14FYP frontier fields, the prediction resolves true. If quantum is demoted below the frontier-field tier in the final NPC text, or if joint NDRC and MOST allocations on quantum information are explicitly cut versus the 14FYP baseline, it resolves false. That is a clean falsifiable bet at the state-laboratory magnitude band — the same kind of bet the other six predictions are, at a smaller capital scale.

Two source pulls, one constant update, one new prediction row

The fix on the date side is mechanical. We pulled the NDRC announcement of the 15FYP NPC adoption, confirmed the date on the NPC archive bulletin, and updated the adoption-date anchor in the pipeline from 15 March 2027 to 12 March 2026. The page renderer reads the anchor and re-derives every resolve-by as the anchor plus the per-prediction lag in months. All six live cards updated their resolve-by dates from 2029 to 2028 and from 2030 to 2029 with no further code changes.

The fix on the quantum side is a new prediction row added to the live-predictions registry, with the plan-line text drawn verbatim from the 14FYP Chapter 9 frontier-fields paragraph and an audit note appended inline. Ranked beneficiary sectors: quantum information (0.70), quantum networking (0.55), quantum metrology (0.38), post-quantum cryptography (0.28). Subsidy band: 1–5 billion USD-equivalent, labelled explicitly as state-laboratory scale with no industrial-fund analog. Lag point estimate: 6 months, with an 80 percent confidence interval of 3 to 12 months — faster than the other predictions because the chapter-level designation lives in the NPC text itself, not in the ministerial follow-on. Falsification: chapter-level demotion below the frontier-field tier, or explicit deprioritization in joint NDRC and MOST quantum-information budget allocations versus the 14FYP baseline. Headline confidence: 0.60, with an 80 percent confidence interval of 0.40 to 0.78, labelled medium.

Three companion source packs landed in the same sprint: a Chinese-original-and-URL pack for five cross-plan excerpts, a 36-entry ministerial-follow-on evidence pack, and a 23-channel monitor pack for the 15FYP rollout. These are the primary-source artefacts the next release's bilingual retrieval substrate will index against.

Six dates shifted, one prediction added, three source packs pinned

The six live-prediction resolve-by dates all shifted twelve months earlier on the page. Sovereign AI and climate-energy moved from 15 September 2028 to 12 September 2027; EV consolidation and embodied AI moved from 15 March 2029 to 12 March 2028; semiconductor self-sufficiency moved from 15 September 2029 to 12 September 2028; common prosperity v2 moved from 15 March 2030 to 12 March 2029. The first resolution window now closes in September 2027 rather than September 2028 — a year earlier than the founding release implied.

The seventh prediction, quantum continuation, is logged with the same schema as the other six, with a resolve-by date of 12 September 2027 computed against the corrected anchor. The live-prediction count moves from six to seven on the page.

The three source packs — Chinese-original verbatim, ministerial evidence, and a forward monitor — are pinned to the project repository and serve as the input substrate for the next retrieval upgrade. Without them, a bilingual sentence-transformer retriever has nothing to embed beyond the existing English-paraphrase corpus.

What a twelve-month error tells us about our own pipeline

The substantive lesson from the date fix is that a forecasting program with a public scoreboard needs to cross-check its anchor dates on every pipeline run, not just at first publication. The original anchor lived in a constant that was correct on the day it was written and silently wrong every day after the NPC session adopted the actual plan. The right pattern is a per-run manifest that records the anchor date the run used and a verification step that pulls the NDRC announcement and asserts the constant matches. The next release wires the propagation; a later release wires the automatic cross-check.

The substantive lesson from the quantum reinclusion is that the absence of an industrial-fund analog is the wrong exclusion criterion for a cascade that compiles into state-laboratory spending rather than industrial-fund spending. The original exclusion was reading the cascade-form question (is there an industrial fund?) when the correct question is the cascade-continuity question (does the chapter-level designation persist?). Re-including quantum at the state-laboratory magnitude band is the cleaner methodological position: the band schema already has a 1–5 billion bucket and a sub-1 billion bucket, and the right call is to use them rather than to exclude the cascade entirely.

What this correction does not claim

The quantum reinclusion adds a seventh prediction; it does not resolve the broader open question of whether other sectors previously excluded (biotech as continuation, defense as guardrail) should be reopened on the same continuity framing. The current position holds biotech as continuation-only on the grounds that the marginal prediction beyond the 14FYP trajectory on the National Reimbursement Drug List is hard to falsify cleanly; that argument may or may not survive the same kind of audit pass we applied to quantum. A later release will revisit the biotech exclusion explicitly.

The date fix is propagated to the page but not yet to the Python pipeline as of this writing — the prediction registry still carries stale dates computed against the old anchor. The next release closes that propagation gap. We surface the disposition here rather than burying it because the principle matters: a public scoreboard with two sources of truth that silently disagree is worse than one that doesn't exist.