Eight working papers documenting the architecture, methodology, and design decisions behind GameChangers since inception — published by Assemble Teams R&D for the community of domain experts this platform serves. These are engineering and design publications, not peer-reviewed research; empirical claims cite primary sources.
Why the Trust Score is computed, never inferred
Architecture of the 0–1,000 composite: six verifiable components (Identity 100, Professional 250, Research 200, Federal 150, Activity 150, Network 150), the auditability requirement that rules out machine-learned scoring for federal and enterprise contexts, and the standing rule that security clearances are self-certified and labeled — never platform-verified.
One constraint, ten uniforms
Our working thesis, tested against primary sources across six sectors in 2026: interconnection backlogs (LBNL), operator succession (EPA/GAO), examiner capacity (GAO/FAA), health-program staffing (WHO/World Bank), agricultural specialists (USDA/DOL), and monitoring workforces (GAO/NOAA) are one pattern — fractional, verification-heavy, experience-gated demand that hiring markets structurally underserve.
Matching AI workloads to jurisdiction and risk
Design of our routing engine: reasoning and contract analysis, structured output, large-document processing, live research, air-gapped contexts, and EU data residency each route to the model class fit for the constraint — with the standing rule that no model output constitutes a compliance determination.
Paying for outcomes in high-stakes work
The mission-contract lifecycle (draft through completion or dispute), escrow locked at signature and released per approved milestone, the time-stamped audit trail (party, timestamp, IP, method), and why outcome-locked payment aligns incentives that hourly billing structurally cannot.
The regulatory boundary we designed on purpose
Framework for surfacing regulatory context — ITAR/DFARS for defense-adjacent work and civilian regimes across Tier-1 sectors (HHS/FDA, DOE/FERC, EPA/SDWA, USDA/FDA, EPA/CAA, DOT/PHMSA) — while the compliance decision remains with mission owners and counsel. Includes the attorney-review gate before any public sector-specific compliance claim.
Why 1,000 members per sector, per country
The gating rationale: marketplaces need dense starting nodes before broad scale; community rooms open only at threshold, seeded with operator-grade prompts; and the structural rule that no fabricated counts or simulated activity ever appear — with the content-integrity incidents that hardened this into policy.
Source-mix, contested tags, and the blindspot
Our publishing methodology: every statistic tied to a named primary source with a methodology note; institutional disagreements displayed rather than smoothed (consensus/contested tagging); analysis always labeled as analysis; and the “blindspot” discipline — naming what mainstream coverage of each sector systematically misses.
AI representation with human authority retained
Post-launch agent architecture: five explicit permission gates (auto-respond, calendar, contract vetting, updates, invitations), permanent action logging, the Delegated Agency Clause in every contract, and the sequencing rule — no agent layer until the human marketplace runs with real users.
Each paper documents work that exists — architecture designed, methodology in use, rules enforced on this site today. Where a capability is post-launch (RD-08), the paper says so. Full technical texts are being prepared for publication as the founding cohort grows; abstracts above are current as of July 2026. Questions or challenges from practitioners are welcome — that's what the community spaces are for.
Reserve your founding place and shape what these papers become.