Category operating system
Launchpad
A versioned product-management source toolkit
A versioned product-management source toolkit with 336 SKILL.md files in the reviewed snapshot, plus orchestration, evaluation, examples, APIs, MCP and platform-specific surfaces. It is an inspectable working environment, not evidence that every method statement or generated artefact is automatically correct.
Business ↔ technology bridge · Category operating system
Turn product judgement into an operating record without outsourcing authority to a model.
Makes full-lifecycle product judgement versioned, inspectable, and executable.
This is a portfolio and source interpretation—not a customer, deployment, certification, or outcome claim.
- 01
Business pressure
Customer, portfolio, economic, delivery, and technical decisions are dispersed across documents and tools, so their original evidence and invalidation logic disappear.
- 02
Product decision
Define which product decisions matter, which evidence advances a gate, who owns the call, and what must cause the decision to reopen.
- 03
Technical record
Versioned skills, the Operated Product Decision Graph, Release Case boundaries, CLI and MCP surfaces, workflow state, evaluations, and audit records.
- 04
Evidence return
Changed assumptions, technical limits, release evidence, and field signals return to scope, economics, sequencing, and portfolio commitment.
Source evidence
What the source actually establishes
A versioned product-management source toolkit with 336 SKILL.md files in the reviewed snapshot, plus orchestration, evaluation, examples, APIs, MCP and platform-specific surfaces. It is an inspectable working environment, not evidence that every method statement or generated artefact is automatically correct.
- Maturity
- Source toolkit under active development
- Evidence review
- Source snapshot through
Architecture and product decisions
- Compile reusable product practices into explicit skills, rubrics, and workflow stages.
- Keep artefact generation, quality evaluation, memory, and trust records as separate concerns.
- Treat Physical AI release readiness as an assurance verdict with human gates, not an automated deployment permission.
Inspectable evidence
- 331 core skill files plus five agent and retrieval packs: 336 SKILL.md files at the reviewed commit.
- Python CLI, API and MCP surfaces with workflow, evidence, audit and quality-test suites.
- A content-addressed HC-OPDG-001 0.1.0 candidate plus reference artefacts and selected evaluation fixtures.
Claim boundary
- Skill inventory and structural depth are source scope, not proof of equal factual authority across every method or model.
- Platform and tool surfaces require path-specific configuration and verification; source presence does not establish identical behavior or external-system access.
- The OPDG bundle remains a candidate with licence and owned-evidence status pending; generated product, safety, legal and release decisions require accountable human review.
Repository evidence is not client, deployment, certification, or outcome evidence.
The Problem
The Problem
Why product teams can't keep up
PM Bandwidth Bottleneck
Product managers can lose substantial time recreating PRDs, stories, roadmaps, and planning material instead of testing assumptions and aligning decisions.
Inconsistent Quality
Artefact quality varies when teams lack shared frameworks, rubrics, and reusable examples. Important reasoning can remain implicit and difficult to review.
Tool Fragmentation
Product context is often distributed across planning, design, engineering, and communication tools. Each hand-off creates a risk that assumptions and evidence lose their traceability.
Scaling Without Hiring
Growing teams need reusable product discipline without assuming that generated artefacts can replace accountable product leadership.
Capabilities
Core Capabilities
10 pillars of AI-powered product management
Source scope only: an entry can describe implemented code, a documented adapter, or a configured workflow. It is not a statement that a production service is active.
01
Expert Skill Library
331 core skills across product, methodology, team, pipeline and Physical AI families, plus five agent and retrieval packs. The structural audit checks required sections and files; it does not establish that every method statement is current or equally authoritative.
02
Virtual Product Team
The source represents 10 virtual role prompts — analyst, architect, developer, UX designer, copywriter, tester, security, DevOps, scrum master, and business analyst. A role prompt is not a credentialed specialist and its output requires accountable review.
03
Berkeley Innovation Pipeline
A repository-defined five-stage pipeline labelled Discovery → Insights → Architecture → Experiment → Deploy, with PM approval gates and fast, standard and careful modes. The label records its declared inspiration; it is not an official UC Berkeley method or endorsement.
04
Workflow Engine
Six source-defined skill chains cover discovery-to-PRD, GTM launch, churn-to-retention, strategy-to-roadmap, sprint cycle and new feature. Context passing is implemented in source; output coherence and factual quality still require path-specific evaluation.
05
Agent Orchestration
The PM Company OS source sequences model-assisted roles, PM approval gates and versioned artefacts. It supports drafting and review; it does not autonomously build, ship or approve a product, and improvement claims require measured evaluation.
06
PM Tool Integration
The source includes manifests and adapter surfaces for Notion, Linear, Jira, GitHub, Figma, Miro, Slack and Confluence. Availability, read/write scope and approval behaviour vary by adapter and require credentials, configuration and path-specific verification.
07
Company Knowledge RAG
The source includes ingestion and retrieval paths for approved playbooks, transcripts and research, with evidence-quality and staleness fields. Retrieval does not establish source correctness, permission, completeness or answer accuracy.
08
Quality Assurance
Source-controlled quality checks, selected evaluation rubrics, and benchmark artefacts make parts of the output reviewable. Coverage and model-judge reliability remain bounded.
09
Multi-Platform Native
The repository contains platform-specific surfaces for a Claude Code plugin, Claude Desktop MCP, ChatGPT actions, Mistral agent prompts and a custom HTTP API. Each path has its own setup and must be verified independently; source presence is not a promise of identical behaviour.
10
Deployment Design
The repository contains container, orchestration, cloud-template, audit, and service-level design surfaces. Templates are not evidence of a production deployment, security posture, or compliance status.
Tech Stack
Built With
Named technologies identify repository dependencies or adapter surfaces; they do not assert a live integration or deployment.
- Backend
- Python 3.12+, FastAPI, Click CLI, Redis caching, Pydantic v2, litellm multi-provider abstraction
- AI / LLM
- Provider-adapter and routing source for Anthropic, OpenAI, Mistral and local-model paths, plus retrieval components; model availability and behaviour depend on configuration and current provider terms
- Frontend
- React 18, Vite, Tailwind CSS, skill browser UI, brief input, output panel, history panel
- Integrations
- Notion (pages), Linear (issues), Jira (tickets), GitHub (releases), Figma (specs), Miro (boards), Slack (digests), Confluence (wikis)
- Infrastructure
- Docker Compose (5 services), Kubernetes Helm, AWS ECS Fargate, GCP Cloud Run, nginx reverse proxy, API key scoping
- Quality & Evaluation
- Pytest structural skill audit, selected rubric and evaluation fixtures, benchmark harnesses, and property-based tests; factual method authority requires separate source assurance
Pricing
Get Started with LaunchpadOS
Contact me to evaluate where LaunchpadOS can support your product practice, what still requires human product leadership, and which integrations would need separate verification.
Request a Demo
Inspect the dated skill inventory, virtual-role and workflow source, and the Berkeley Innovation Pipeline implementation. The demo separates what is present in source from what is configured, tested and ready for accountable use.
Physical AI Product Leadership Mission
Use the toolkit inside an accountable product mission.
Hyperion can use selected LaunchpadOS source patterns while leading one evidence-backed product milestone. The commercial mandate remains the Physical AI Product Leadership Mission; LaunchpadOS is supporting R&D, not a fourth offer or a promise of autonomous delivery.
Explore the Product Leadership MissionThe AI Dispatch
From demo to dependable operation. Get a weekly decision note for Physical AI products.
Unsubscribe anytime. No spam, ever.