Post-AGI Planning

A living, data-driven civilization forecast and long-form guide to the post-AGI transition. The REHOBOAM timeline spans 2026–2040 across six domains and is published at peterxing.com.

Every individual prediction must have exactly one reviewed direct X evidence card. The matcher prefers a defensible post/repost observed in @peterxing's activity, then retains a reviewed authoritative external post labeled as direct, scenario, or leading-indicator evidence. Source freshness is tracked separately from evergreen evidence age. refresh-signals.js exits nonzero and leaves signals.json unchanged unless direct coverage is complete and signals.search is empty.

How it works

index.html ──loads──> styles.css + app.js
           ├─fetch──> predictions.json   (forecast source of truth)
           └─fetch──> signals.json       (generated X matches + Reality Signals)
                              ^
                              |
                  refresh-signals.js ──> x-archive.js
                                           ├─ Wayback CDX ID discovery
                                           ├─ X first-party tweet-result hydration
                                           └─ X oEmbed authorship cross-check

• predictions.json contains the probabilistic 2026–2040 forecast.
• validate-predictions.js enforces schema plus portfolio coherence: exact/near duplicates, repeated valuation/science/AI-R&D endpoints, conventional career milestones after full automation, and unlabelled top-expert milestones after an earlier ungoverned ASI branch all fail publication. Similar later events must advance a threshold, scope, deployment stage, or branch.
• refresh-signals.js expands the forecast into one matcher per event. Literal scoring is supplemented by a bounded concept ontology for semantically equivalent evidence (such as tape-out/semiconductors, physicians/health, FSD/robotics, and UHI/dividends). Claim-specific facet guards remain mandatory for literal, semantic, hybrid, and family matches. Assignment maximizes unique reviewed posts first, then permits reuse only inside a declared compatible evidence family or reviewed threshold/scenario series. Publication fails when one status supports more than 10 predictions.
• x-archive.js discovers Peter activity IDs through fully paginated Wayback CDX queries for both twitter.com/x.com and both handle case variants. It merges API-era private history and public historical signal bundles, numerically stratifies IDs across 2015–2026, hydrates through X's first-party per-status endpoint at least 600 ms apart, cross-checks through X oEmbed, and stores the deduplicated authored/quote/reply/repost corpus only under pap-secrets.
• evidence-families.js declares the only families within which threshold-series reuse is compatible. Cross-family reuse fails publication.
• evidence-approvals.json is the public-safe, sticky reviewed prediction/post-pair ledger. Each approval is bound to the exact prediction text and retains public provenance plus review and verification dates. New automatic candidates cannot self-approve, and publication fails below the reviewed 24-mapping Peter floor and 10-mapping Peter-authored floor.
• external-evidence.js is the reviewed authoritative-source ledger. It stores only public-safe status metadata, source-quality classification, scenario/leading-indicator labels, rationale, and compatible reuse groups.
• verify-signal-matcher.js runs positive and negative regression fixtures for the concept matcher, including J-space vs off-world space, building permits vs compute permits, market cap vs compute caps, political vs electrical power, and quantitative labor thresholds.
• signals.json is written only at complete direct-only coverage. signals-debug.json remains local and records source freshness, historical span, missing direct IDs, reviewed mappings, source-failure class, guard rejections, and reuse audits without storing the raw activity corpus.
• index.html, styles.css, and app.js form the cacheable static shell. Forecast data is not duplicated inline; a missing sidecar produces an explicit unavailable state.
• author.json drives the daily-refreshed About the Author section.

Data-source safety

Bulk profile retrieval is not a publication dependency: the former nitter/RSS and legacy syndication paths are unavailable, and X API quota/auth/plan failures are diagnostic only. The durable chain is:

1. Wayback CDX activity-ID discovery, fully paginated across both hosts and handle case variants.
2. Existing private API-era history and public historical signals.json IDs.
3. X first-party tweet-result hydration for public status IDs, paced at least 600 ms.
4. X oEmbed cross-check of original authorship and Peter's activity relationship.
5. Manual prediction/status approval with strict concept and facet guards.

Reviewed evergreen evidence age is separate from the current verification sweep. Every publish re-hydrates all selected original statuses and cross-checks their authors; deleted/protected records fail closed. signals.sourceStatus, signals.sourceAttempts, and the UI expose the archive-verified chain and any X API diagnostic failure.

Raw activity and credentials stay outside the repository.

Local use

npm install
npm run validate
npm run refresh
npm run verify:matcher
npm run serve

In a second terminal:

npm run verify
npm run verify:predictions
npm run verify:reality
npm run verify:author
npm run verify:ui
npm run verify:performance
npm run verify:coverage
npm run verify:archive
npm run verify:peter
npm run verify:external

npm run refresh:archive advances a cache-aware 120-status verification batch. X_ARCHIVE_BACKFILL=1 X_ARCHIVE_HYDRATE_LIMIT=400 node refresh-signals.js imports historical public signal versions and advances a larger, numerically stratified batch. X_ARCHIVE_DISCOVERY_FORCE=1 refreshes every Wayback page. X_SKIP_API=1 proves publication does not depend on authenticated X. For reviewed external-source research, node x-archive.js --account=HANDLE --hydrate=120 uses the same private Wayback/first-party/oEmbed chain without adding anything to the public ledger. npm run review:candidates emits ID-only, strict-guard candidate diagnostics; inspect an individual public post with node review-evidence-candidates.js --show-public=STATUS_ID before editing an approval. Candidate generation never edits either evidence ledger.

Deploy

powershell -ExecutionPolicy Bypass -File .\deploy.ps1

There is no build step. Vercel serves the split HTML/CSS/JS shell and JSON sidecars as static files.

Security

• Credentials are loaded from C:\Users\peterxing\pap-secrets\.env and are never served or committed.
• Raw caches, debug output, logs, .vercel, and node_modules are excluded from GitHub.
• server.js is default-deny and serves only the site, public JSON sidecars, and static assets.

See X-API-SETUP.md for X authentication and REVISE-PREDICTIONS.md for forecast revision rules.