Knowledge Map
Extracted knowledge organized by domain. Every entry traces back to one or more sources with quotes, context, and links. See meta/taxonomy.md for classification rules.
Entry types: technique · mental-model · insight
Entry status: draft → solid (multiple sources) → canonical (vetted, teachable)
Product Lifecycle
Six phases. Entries sit flat within phase folders; the component frontmatter field records the most specific applicable component. See meta/lifecycle-framework-v2.md for full phase lineage and AI-PM emphasis.
Discover
What problems are worth solving? Components: Problem Signal Detection · Market & Competitive Intelligence · Opportunity Prioritization · Problem Brief
No entries yet.
Frame
What does success look like, and what’s the bet? Components: Stakeholder Intent Alignment · Success Criteria & Business Case · Roadmap Definition
No entries yet.
Shape
What solution takes form, and for whom? Components: Persona & Journey Mapping · Prototyping & Risk Reduction · Go-to-Market Planning
- Shape the Product — AI Disruption Profile — Determining what to build; most disrupted PM job (🤖🤖🤖), especially strategy/vision (🤖🤖🤖🤖)
Build
How do we ship with clarity and conviction? Components: Feature Specification Writing · User Story & Acceptance Criteria · Stakeholder Communication · Scope & Priority Tradeoffs
- Interactive PRD Writing — Templatized rule files + AI follow-up questions for thorough PRDs
build/feature-specification-writing - Task List Generation for Observability — Decomposing PRDs into nested task lists for observability and control
build/user-story-acceptance-criteria - Context First Development — The biggest mistake in AI-assisted dev is rushing past context
build - Ship the Product — AI Disruption Profile — Helping the team deliver; moderately disrupted (🤖🤖), scope tradeoffs remain human-driven
Release
How do we put this into the world deliberately? Components: Release Readiness Assessment · Phased Rollout Strategy · Release Communications · Launch Marketing & Enablement
No entries yet.
Measure
Did it work, and what do we do next? Components: KPI & Outcome Monitoring · Customer Feedback Collection · Experiment Design & Analysis · End-of-Life & Deprecation
No entries yet.
Horizontal Domains
Lifecycle-agnostic AI PM skills and knowledge areas, organized by delivery mechanism with increasing capability and autonomy. See taxonomy for classification rules.
Prompting — Portable Techniques
Portable techniques for crafting effective instructions — works in any chat window. Prompting patterns, meta-skill patterns, writing workflows, role delineation.
- Be 100x More Specific — Forces AI past vague principles into concrete, actionable standards
- My Job Your Job Role Delineation — Explicit human/AI responsibility partitioning
- AI as Writing Coach — Structured workflow: thesis validation → blind spots → restructuring
Context & Knowledge Management — Knowledge Infrastructure
Making non-code knowledge discoverable and usable to agents and their human coworkers — context graphs, agent-oriented knowledge management, progressive disclosure.
- Deliberate Context Selection — Hand-picking files for LLM context vs. relying on automatic context
- Sync the People — AI Disruption Profile — Human coordination and alignment; least disrupted PM job (🤖), a durable competitive advantage
- Three-Layer Context Disclosure — Index → summary → full content: the converging pattern for efficient agent retrieval (~10x token savings)
- Filesystem as Retrieval Architecture — Directory hierarchy as index, frontmatter as metadata, git as temporal layer — a legitimate retrieval system, not a stopgap
- Repositories as Context Boundaries — Repos shift from code isolation to context containers; git subtree/submodules as cross-repo context distribution; agents absorb submodule ceremony
Templated AI Runtimes — Packaged AI Tools
Packaged, shareable, non-agentic AI tools (Custom GPTs, Google Gems, Claude Projects). Building, distributing, and managing templated AI runtimes for teams and organizations.
No entries yet.
Agents — Building & Managing Knowledge Agents
Building and managing knowledge agents — lifecycle management, rules, skills, templates, tools, workflows. How PMs select, onboard, train, and performance-manage AI agents.
- Stepwise Task Execution — One-task-at-a-time execution with pause-and-approve checkpoints
- Meta-Skill Pattern — Build a “skill that builds skills” to bootstrap agent capabilities consistently
- Filesystem as Agent State — Agent architecture = filesystem (state) + LLM (orchestrator); company-as-filesystem gives agents a shared namespace
- Knowledge Capture as Side Effect — Design agent systems so knowledge capture is a byproduct of corrections, not a separate task; extends to self-driving documentation
solid - Agent as Cross-Tool Workflow Hub — Local agent + MCP integrations replaces manual cross-tool workflows; becomes the orchestration layer across disconnected SaaS tools
- Agent-Mediated Self-Reflection — Using agents to observe your behavioral patterns (conflict avoidance, intention-action gaps, unregistered learnings) from digital exhaust
- Progressive Tool Disclosure — Revealing MCP tools in layers to combat choice paralysis and hallucination (+15% accuracy vs flat exposure)
AI Adoption & Change Management
How organizations and individuals adapt to AI-native ways of working — scaling expertise, restructuring teams, driving adoption.
- Reverse Engineer Judgment Into AI — Have AI discover your implicit criteria, encode into reusable evaluator
- Scale Manager Expertise With AI — Automate “0-to-60%” feedback so managers focus on high-leverage work
- AI Disrupts Strategic PM Skills Most — Counterintuitively, AI most disrupts high-level strategic PM skills, not soft skills
- Data Silos Block Enterprise Agent Adoption — Enterprise data fragmented across SaaS tools is the primary barrier to agent rollout, not model capability
- Tool Identity as Adoption Gate — When a capable AI tool has narrow adoption, the bottleneck may be naming/branding/positioning, not capability
- Retrieval Infrastructure Graduation — Tiered path from filesystem-only → semantic search → knowledge graph, with graduation criteria for each transition
Software Methodology Evolution
How AI fundamentally changes the way software gets built — compound engineering, spec-driven development, vibe coding, evolving delivery paradigms.
- Spec-Driven Development — Software ships as specs + tests with zero code; works for stable utilities, breaks for anything needing performance, debugging, community, or security patching