★ One-page overview (start here)
★ The engine, explained (for everyone)
· Overview
The thesis in one page — Playroll as a coverage-directed data engine for world models, not a data brokerage and not a frontier lab. The annotated flywheel with evidence grades per arrow, the strategic position, and how this series relates to the quests thread.
· Market & Buyers
Who buys world-model training data as of mid-2026, what the pricing comparables actually are, what buyers require, and the five places the demand thesis is weakest. Every number traced to a primary source in the research pass.
· Technical Evidence
Why action-aligned data matters (and where it doesn't) — real vs latent action labels, diversity vs volume, the honest games→robotics read, and the state of neural game engines. Each claim graded strong / suggestive / speculative.
· Competition & Moat
The competitor map (Medal/General Intuition, Grunt, Overwolf, NVIDIA, platform owners, marketplaces), what each actually captures, how fast each could replicate inputs+quests, and the honest 12-month vs 3-year moat assessment.
· Engine Architecture
The coverage-directed data engine, component by component — eval & attribution layer, quest compiler, fleet capture, curation & QC, consent ledger — with the prior art each borrows from (Tesla triggers, Waymo rare-example mining, MATES, Llama-3 annealing, VPT, Ego4D) and the whys behind every design decision.
· Products & Publishers
The three products in order of shipability — directed-collection-as-a-service, the Generalist-IDM and curated slices, and studio playtesting agents — plus the publisher flywheel that converts the
· Risk & Compliance
The ranked risk register — publisher IP as risk
· Roadmap & Playbook
7 elementi
· AI Delegation
The living map of which data-engine work is delegated to AI agents — tier definitions, the delegation rules, the full work-item map with the data each step needs, and the iteration loop that keeps deepening it until everything delegable is delegated.
0 · Agent Briefs
Dispatch-ready briefs for the first batch of AI-delegated work — EULA audit, POC eval harness, coverage atlas, benchmark probes, curation pipeline, IDM scaffold, provenance manifest, and fraud-forensics features — each with inputs, deliverables, acceptance tests, and escalation triggers.