Skip to content
SynthWorldDeterministic identity benchmarks
Esc
navigateopen⌘Jpreview
On this page

SynthWorld roadmap

This file records direction rather than a release promise. The user-facing Now/Next/Later view lives at https://bluntmachetti.github.io/synthworld/roadmap/; governed registries remain authoritative for current capability and benchmark status.

Product principles

Ground truth first; public observations stay separate from evaluator truth; synthetic-data safeguards remain mandatory; replay inputs and versions stay explicit; small frozen fixtures remain stable while larger workloads are generated separately.

Architecture direction

SynthWorld remains a deterministic identity/evaluation layer rather than a general runtime. Domain systems consume public projections and return predictions or traces for independent scoring.

Phase 1 — Benchmark adoption

The shared evaluation framework and core benchmark families are established. Current work is documentation clarity, publication discipline, and reproducible benchmark consumption.

Phase 2 — World depth and longitudinal truth

Richer graph profiles and broader deterministic temporal composition remain active directions under #3 and #2 while existing frozen fixtures stay stable.

Phase 3 — Priority market packs

Priority work builds on the shipped broker lifecycle and agent-authority surfaces while generated scale and deeper composition continue separately.

Data-broker deletion and reappearance

Broker lifecycle evaluation is shipped; broader longitudinal product behavior remains separately scoped.

AI agents and non-human identities

Frozen agent-authority conformance surfaces and the configurable generated enterprise-agentic tiers are shipped: smoke retains its V1 contract, while standard and longitudinal use the independently versioned V2 scale and lifecycle contract. The generation and evaluation paths being available does not make the governed enterprise-agentic benchmark family published; its registry lifecycle remains authoritative. Stress-scale worlds, broader deterministic composition, and large-world Explorer workflows remain separate future work.

Phase 4 — Portfolio and AI-system integrations

The enterprise identity/access foundation and contextual/continuous assurance contracts are established. Remaining work extends those bounded contracts without turning SynthWorld into a general enforcement product.

Phase 5 — Exploratory identity ecosystems

Later exploration includes LLM/RAG privacy (#8), digital wallets and verifiable credentials (#9), and disaster identity continuity (#10).

Use-case map

Use the documentation site for journey guidance and the generated benchmark catalogue for governed current state.

Explicit non-goals

SynthWorld does not anonymise supplied real-world data, impersonate real people, replace domain runtimes, or turn deterministic benchmark scores into forecasts.

Contribution guidance

New benchmark packs should begin with an issue that defines public input, evaluator truth, versioning, deterministic inputs, metrics, negative controls, and the frozen-versus-generated artifact strategy. See CONTRIBUTING.md.

Was this page helpful?