Docs/roadmap

Roadmap

Status: v1.0 corpus · community adoption Draft
Product: AIES — AI Engineering Standard
Updated: 2026-07-16
Repository: grvmithas/open-ai-engineering-standards


Purpose

This roadmap separates two things teams confuse:

  1. Engineering corpus — do we already know how to run production AI delivery?
  2. Community standard adoption — have external orgs validated, forked, and governed this together?

AIES already ships the corpus. The remaining roadmap is adoption, packaging, and governance — not inventing prompts, MCP, or AIDLC from scratch.

What is already shipped (corpus maturity)

Treat these as present, not “planned learning”:

Domain Where it lives
Prompt engineering & XML contracts 01-prompt-engineering/
Context engineering, packing, progressive disclosure 02-context-engineering/, handbook/how-to-give-context.md
Skills vs prompts vs agents handbook/skills-vs-prompts-vs-agents.md, 03-skill-engineering/
Agent engineering, debug/planning skills 04-agent-engineering/, patterns + handbook
Multi-agent / orchestration 05-multi-agent-systems/
Memory, RAG, caching, poisoning defenses 06-memory-knowledge/
Evaluation, LLMOps, judges, gates 07-llmops/, handbook/evals-as-quality-gates.md
AIDLC / AI SDLC, HITL, modern coding vs vibe coding 08-ai-sdlc/, handbook planning/HITL/PR-review
MCP when/how + reference integrations handbook/mcp-when-and-how.md, mcps/
Hooks, tokens, cost discipline handbook chapters + LLMOps
New AIDLC roles & org operating model 10-ai-org-playbook/, handbook/aidlc-roles-and-evolving-delivery.md
Patterns (planner, reviewer, release, hotfix, …) patterns/
Normative controls, SoA, conformance tooling standard/, portal, schemas
Platform cookbooks (adapters) cookbook/

Adoption track (what “next” actually means)

Track Theme Exit criteria Status
A1 Corpus publish & site Curated nav; handbook; portal; principles; honest positioning Done
A2 Packaging & discoverability Stable URLs; CONTRIBUTING usable by strangers; search index trustworthy In progress
A3 External pilots ≥3 org pilots with published notes (what failed, what stuck) Next
A4 Shared governance RFC process live with non-steward reviewers; MAINTAINERS expanded Planned
A5 Inter-assessor reliability Two independent teams apply same SoA sample; disagreement notes published Planned
S1.0 Community standard claim A3–A5 met; no portfolio-hero positioning; versioned normative freeze Future

Do not read A3–A5 as “we still need to invent MCP or evals.” Those practices are already in-repo.

Release philosophy

  1. Corpus first — depth of working playbooks beats empty version theater.
  2. Adoption is earned — “standard” means external critique and shared governance, not a badge on the homepage.
  3. Evidence over slogans — pilots, checklists, and failing gates move the track — not page count.
  4. Vendor neutral — no release may hard-require a single model vendor.

How to influence this roadmap

  1. Open an issue or PR against this file with a concrete adoption exit criterion.
  2. Prefer pilot evidence and failing checklists over “add another planned version for X we already documented.”
  3. See Contributing.

Anti-patterns for this roadmap

  • Listing shipped capabilities as “Planned” (misrepresents maturity)
  • Declaring community-standard S1.0 because the docs site looks finished
  • Adding 100 empty pages to inflate “coverage”
  • Pinning the standard to one vendor’s API surface

Changelog

  • 2026-07-16: Reframed roadmap — corpus shipped; adoption track replaces false “planned invent” versions.
  • 2026-07-16: Initial Draft framing published (superseded).