ADR-0010 — Adoption routing (soft-with-teeth): doom bends where work is adopted

Status: ACCEPTED · Decided: 2026-07-12

These are the actual decision records the game is built from, published unedited except for internal process notes. They describe why the game works the way it does. They are not a strategy guide, and some of them argue with each other.

Status: ACCEPTED
Date: 2026-07-12

Context

The exploit sweep's dominant line (safety_lean, basement safety-spam, median 37 turns vs 7 for passive) exists because safety research bends doom directly, privately, and state-independently — papers are decorative, conferences float disconnected. Pip's paper taxonomy (this session): papers raise awareness, build the writer's reputation, speed follow-on work, and get adopted — into government policy or frontier-lab practice, including rivals'. Pip's normative ground, verbatim:

"people voluntarily adopting good actions will generate the space for better actions to come about, and this is how small and middle power players can affect things — norm and market setting"

Decision

Soft-with-teeth routing:

Your own lab's doom contribution is basement-fixable. Private safety work makes your lab safer, no adoption required. (Honest: you really can secure your own shop in private.)
World/rival doom — the majority share (#562 rival scaling) — bends only through adoption: research → paper → socialization (conferences, summits, direct stakeholder work) → adoption by labs and/or governments → doom bends. Unadopted world-facing research is a PDF.
Reputation is per-person and per-org, and attention is typed: safety successes attract interviews/grants/safety applicants; reckless capability moves attract accel VC money. Both directions are priced, neither is forbidden (the dual-use temptation is a design feature — see DESIGN_PHILOSOPHY, tension/tradeoffs).
Papers additionally speed follow-on work (tech-tree effect) and unlock mitigations.
Event-horizon guardrail (lore): the player's influence over other labs becomes substantial mostly after the run crosses the real-world event horizon — the game never implies real labs do the truly weird late-game stuff. (WORLD_AND_LORE.)

Beacons served / violated

Rams #6 (honest): the mechanic is Pip's real-world theory of change — norm and market setting by small/middle powers. Structure, not numbers; don't patch it for freshness (per the philosophy doc's load-bearing-structure rule).
MaRo Interaction: research, reputation, conferences, and rivals now interlock; the social layer stops being decoration.
Kills the dominant line by construction: basement-spam can't reach the dominant doom mass, so no payoff buff to safety-spam can restore it.

Interaction contract

Reads/writes: rival-doom pipeline (#562 — rival share must dominate the integral for teeth), reputation (per-person + per-org, new read/write), conference/travel (promoted from flavor to load-bearing — the socialization step), Liability Ledger (funding-with-strings, promises), SA (awareness essays raise it), exploit-finder (new sweep target: publish-and-socialize must beat basement-spam).

Rejected alternatives

Strong routing (ALL safety value requires adoption): less honest — you can privately secure your own lab — and it needlessly breaks the existing own-lab safety architecture.
Status quo (direct doom reduction): sweep-proven degenerate.

Consequences / open questions

Safety research splits: internal hardening (own-lab, private) vs publishable agendas (world-facing, adoption-routed).
Conference/travel design (agenda item 4) is promoted: it's now the mandatory middle of the value chain, not garnish.
Balance verification target for the next sweep: rival/world doom share must dominate the integral, or the teeth are cosmetic.
Adoption by whom needs a counterparty model — overlaps ADR-0007 alliances and DQ-9 receivables; design together.