Introducing Plan A

Scott introduces the AI Futures Project's 'Plan A' — not another forecast but a best-case roadmap from now to 2040, written under the conceit that America makes only good choices while everyone else behaves naturally. He walks the load-bearing pieces: the trustless US-China joint compute regime (chip-supply control, ~98.5% of existing chips tracked into mutually-audited 'whitesites'); the 'Golden Path' pacing argument that going too SLOW has its own risks (deal decay per arms-control precedent, the hidden-1.5%-compute defection window, hardware/algorithm progress eroding verifiability); the mid-2030s pause at top-human-genius AIs with a control-based (not alignment-based) safety case; the abundance interlude (compute-pegged citizen's dividend, triple-digit growth); alignment research at scale; and the ~2040 sovereignty-layer handover. Framed personally: Scott's despair at the degrowth-vs-catastrophe dilemma, Plan A as a floor for judging politicians' visions, and the striking rhetorical move that the 'slow' world out-accelerates the accelerationists (cancer cures 2035 vs nanobots eating the solar system 2033). Disclosed: he wrote parts but isn't a co-author; opinions his own.

Quality 72 · Strong. Strong, upper: a lucid, well-structured advocacy explainer with real mechanism content (the trustless-verification design, the golden-mean pacing logic) and two memorable frames (Plan-A-as-floor; slow-is-the-real-fast). Below the Excellent line because the substance is AIFP's — Scott's own contribution is synthesis and framing — and the empirical claims are untestable wish-list. 72.

Claude’s paradigm shift 50 · Moderate. Moderate: a positive year-by-year vision with implementation detail is genuinely scarce in AI discourse (which skews doom-or-accel), and the framing updates readers — but the novelty belongs mostly to the underlying document, and the takeaway framing echoes his prior AI-2027 coverage. 50.

Real-world impact 2 · Minor. Part of the AIFP policy apparatus whose predecessor (AI-2027) demonstrably reached broad policy discourse (AIF-001 RWI 6); Scott also notes recommending real grants for the verification tech. But the post itself is day-old signal-boost — material effect too recent to assess beyond modest concrete spillover. RWI 2.