OpenReason Roadmap

The planned development of OpenReason — from core protocol through tooling, agents, side protocols, and integrations.

This is a living document subject to OpenReason principles. If you disagree with the sequencing or priorities, open a discussion or fork it.

Philosophy of Sequencing

Good open source infrastructure earns adoption in layers:

1. Protocol must be implementable manually before tooling is built
2. Tooling must be stable before agents are layered on top
3. Agents must be trustworthy before integrations embed them in workflows

Each phase must be genuinely complete before the next begins — not feature-complete, but stable enough that contributors can build on it without fear of breaking changes.

Phase 0 — Foundation

Status: Complete Target: Specification and philosophical foundation

The goal: Produce documents coherent enough to invite rigorous challenge before any code is written.

Completed

• ORP v0.1 specification
• REE philosophical foundation
• README and project overview
• CONTRIBUTING guide
• Danish Property Tax Reform (worked example, ORP-Full)
• ORP_META.md (protocol’s own provenance)
• Governance model v0.1

Public comment period: Open for community challenge ongoing

Phase 1 — Core Tooling

Status: In Progress (Sprints 1.1-1.6) Target: v0.5.0

The minimum viable toolkit that makes ORP documents easy to produce, validate, and compare.

Completed (Sprint 1.1-1.5)

• ORP CLI (orp) — Commands: new, validate, check, diff
• ORP Validator (Python package) — pip install orp-validator
• ORP Schema (JSON Schema for ORP v0.1)
• Reference implementation (Danish Property Tax Reform)
• REST API — api.publicreasonproject.org

In Progress (Sprint 1.6)

• Documentation website — docs.publicreasonproject.org
  • Getting Started guides
  • Protocol specification (web-friendly)
  • Tool documentation
  • Examples showcase
  • Community pages

Next (Sprint 1.7+)

• Interactive validator widget (browser-based YAML validation)
• JavaScript/Node validator package
• Additional worked examples

Target completion: March 2026

Phase 2 — Developer Experience

Status: Planned Target: v0.6-v0.8

Making ORP feel native to existing developer workflows.

Planned Features

GitLab Integration

• ORP document templates for repositories
• GitLab CI pipeline validating ORP compliance on merge requests
• Automatic fork registry updates
• Issue templates for critique and amendments

VS Code Extension

• Schema-aware autocomplete for ORP YAML
• Inline validation with error highlighting
• Field documentation on hover
• Compliance level indicator in status bar

ORP Registry

• Public registry of ORP-compliant documents
• Searchable by domain, compliance level, author, topic
• Forks and responses linked automatically
• The public ledger that makes forking visible

ORP Viewer

• Clean web interface for reading ORP documents
• Renders five layers in plain language
• Shows fork tree
• Interface for non-technical stakeholders

Target: Q2 2026

Phase 3 — Agent Layer

Status: Planned Target: v1.0

AI-assisted tooling that dramatically reduces the cost of producing high-quality ORP documents.

Critical Principle

Agents assist, they do not attest.

Every field produced by an agent must be reviewed and signed off by a human before a document reaches ORP-Standard or ORP-Full compliance.

Planned Agents

Provenance Agent (L1)

Given a policy domain and brief description:

• Identifies commonly used datasets
• Researches known limitations and coverage gaps
• Flags likely exclusions
• Drafts L1 with sources cited for review

Most valuable agent — L1 is where most documents are weakest and where bias lives.

Simulation Agent (L2)

Given a proposal and variable definitions:

• Identifies standard modeling approaches
• Suggests affected population segments
• Runs basic scenario analysis
• Flags assumptions requiring human decision

Empathy Mapping Agent (L3)

Given a proposal and affected population:

• Identifies stakeholder groups including non-obvious minorities
• Estimates differential impacts
• Flags groups whose impacts couldn’t be modeled
• Prompts for explicit minority stress-testing

ORP Assistant

Conversational interface for producing ORP documents. Asks the right questions in the right order, explains why each field matters.

Designed for:

• Politicians and policy staff without technical backgrounds
• Students using ORP in educational contexts
• Journalists investigating public decisions

Fork Analyser

Given original and fork:

• Identifies specific assumptions that differ
• Explains in plain language why differences matter
• Estimates outcome differences if fork’s assumptions are correct
• Generates structured debate brief for each position

Makes policy disagreement locatable and productive.

Target: Q3-Q4 2026

Phase 4 — Side Protocols

Status: Planned Target: v1.x extensions

Domain-specific extensions adding fields and methodology standards. All backwards compatible with core ORP.

Planned Extensions

ORP-AI (AI Training Data)

Extends L1 for AI and machine learning:

• Dataset lineage and chain of custody
• Demographic coverage analysis and gaps
• Synthetic data generation methodology
• Model training decisions and data implications
• Fine-tuning data documentation

Addresses core REE insight: Bias lives in data selection and cleaning, not algorithms. Makes that layer auditable.

ORP-Econ (Economic Modelling)

Extends L2 with econometric fields:

• Model specification and identification strategy
• Causal vs correlational claims
• Confidence intervals and sensitivity analysis
• Peer review status of methodology

ORP-Env (Environmental Impact)

Extends L2 and L3 for environmental/climate policy:

• Carbon accounting methodology
• Intergenerational impact modeling
• Ecosystem stakeholder mapping (non-human sentient beings)
• Uncertainty ranges for long-horizon projections

ORP-Health (Healthcare Policy)

Extends L3 with healthcare-specific fields:

• Clinical outcome measures
• Health equity analysis across demographics
• Access and affordability modeling

ORP-Edu (Educational Use)

Simplified implementation for classroom use:

• Same five layers with age-appropriate language
• Guided examples
• Scoring rubric for educational assessment
• Interface between OpenReason and policy simulation games

Target: 2027

Phase 5 — Platform and API

Status: Planned Target: v2.0

Hosted infrastructure making OpenReason accessible without local tooling.

Planned Services

ORP API

REST API accepting policy proposal in plain text, returning AI-assisted first draft across all five layers.

POST /api/v1/draft
{
  "title": "Policy Title",
  "domain": "policy",
  "description": "...",
  "compliance_target": "standard"
}

→ Returns: Partially populated ORP document
          with agent-drafted fields flagged for review

ORP Registry API

Programmatic access to public document registry — search, retrieve, compare, track forks.

Infrastructure for third-party tools to build on.

OpenReason for Education

Hosted platform for schools/universities:

• Classroom management
• Student submissions
• Peer review workflows
• Teacher assessment tools
• Built on ORP-Edu

OpenReason for Government

Hosted enterprise implementation:

• Audit trails
• Multi-party attestation
• Integration with document management systems
• Compliance reporting (e.g. EU AI Act)

Target: 2027-2028

What We Are Explicitly NOT Building

To stay focused:

Current Sprint Status

Phase 1 Progress:

Next: Sprint 1.7.0 — Interactive validator widget

How to Influence the Roadmap

This roadmap is an ORP document. It has assumptions, priorities, and will be wrong about some things.

The right response to disagreement is not a comment — it’s a fork.

• Think Phase 3 should come before Phase 2? Fork and show your reasoning.
• Think ORP-AI should be in Phase 1? Make the case.
• Have a different priority? Propose it.

The roadmap evolves through the same transparent, forkable reasoning OpenReason enables everywhere else.

See Contributing Guide for how to propose changes.

Resources

• Repository: gitlab.com/publicreason/orp
• Issue Tracker: gitlab.com/publicreason/orp/-/issues
• Full Roadmap: ROADMAP.md
• Sprint Plans: docs/sprints/

OpenReason Roadmap v0.1 — February 2026 This document is subject to ORP principles. Fork it.