The Signal Method: How to Build Products That Never Lose Context

Most products are built on amnesia.

A meeting happens. A decision gets made. Someone starts building. Three months later, nobody remembers why that feature exists, what it was supposed to solve, or what was explicitly ruled out of scope.

The code is there. The context isn’t.

This is the problem the Signal Method solves.

What It Is

The Signal Method is an open-source methodology — MIT-licensed — for how SaaS companies should build products. Fork it. Clone it. Make it yours.

It’s not a framework. It’s not a process. It’s a knowledge structure. A single repo that holds everything about how your product is built — not just code, but the decisions, strategy, and context behind the code.

The Pipeline

Every significant feature follows a traceable pipeline:

Transcript → Mission → Vision → PRD → Design → Architecture → Execution → Completion.

Each stage produces a versioned markdown artifact. Nothing is lost. Everything is traceable.

Transcript. A meeting happens. A conversation about a new feature, a problem, a direction. The raw input gets captured.

Mission. Why does this initiative exist? What problem are we solving? Who does it affect? What happens if we don’t solve it? This is the “why” — documented, not assumed.

Vision. What does success look like? Measurable criteria. Explicit scope. Explicit non-goals. “Out of scope” is just as important as “in scope” — it prevents the feature from becoming everything to everyone.

PRD. What must be built? Requirements, acceptance criteria, dependencies, constraints. Not a 40-page document nobody reads — a clear, concise definition of done.

Design. How do users interact with it? Flows, edge cases, system interactions.

Architecture. How is the system structured? Components, data models, integration points, technical decisions.

Execution. The code. PRs linked to the initiative. Commits traceable to requirements.

Completion. What was delivered? Outcomes vs. success criteria. Lessons learned. Follow-up items.

From any outcome, you can trace the full line backwards: Completion → Execution → Architecture → Design → PRD → Vision → Mission → Transcript. The why behind every line of code is documented.

Why Markdown. Why the Repo.

Everything in the Signal Method is a markdown file in a git repo. Not Confluence. Not Notion. Not a wiki that nobody updates.

Why?

Because your most powerful resources need access to it.

AI agents are writing code, reviewing PRs, making architectural decisions. But they operate in the repo. If your strategy, your decisions, your meeting outcomes live in Confluence, your AI agents are working blind. They have syntax but no intent. Code but no context.

When your knowledge lives in the repo — as structured markdown, versioned, traceable — every AI agent working on your codebase can reason about why something is being built, not just how. They can reference the mission when making architectural choices. They can check the PRD when reviewing a pull request. They can understand which initiative is the priority this quarter.

The repo becomes the complete context layer. Code + intent + decisions + strategy. All in one place. All accessible to both humans and AI.

The Four-Layer Intelligence Stack

The Signal Method organizes intelligence into four layers:

Intelligence Layer. Transform raw data into signals. Commits become classified. Activity becomes patterns. Noise becomes signal.

Strategy Layer. Capture intent and evaluate alignment. Initiatives, visions, roadmaps — connected to the execution happening beneath them.

Knowledge Layer. Preserve organizational context. Meetings, decisions, conversations, research. The institutional memory that usually dies in Slack threads.

Reasoning Layer. Synthesize across all layers. This is where AI reasons over the full picture — connecting a meeting decision to a code pattern to a strategic initiative to a cost signal.

Each layer builds on the one below it. Intelligence feeds Strategy. Strategy feeds Knowledge. Knowledge feeds Reasoning. Reasoning produces insight.

Core Principles

Signal Over Noise. Every system produces data. Most of it is noise. The Signal Method demands that you extract what matters before presenting anything. Not “here’s your dashboard with 47 metrics.” Instead: “here’s what changed, why it matters, and what you should consider.”

Truth Over Comfort. The system tells you what you need to know, not what you want to hear. Bad news is shown clearly, not hidden. Uncertainty is quantified, not disguised as confidence.

Intent to Outcome. Every action traceable from intent through execution to outcome. If you can’t trace the line, you can’t reason about it. If you can’t reason about it, you can’t improve it.

AI as Infrastructure. AI isn’t a chatbot sidebar. It’s a reasoning layer that sits on top of your product’s structured data — signals, strategy, knowledge, and code — and synthesizes them into intelligence.

Non-Goals Are Goals. Knowing what you will never build is as important as knowing what you will. It prevents scope creep, maintains clarity, and communicates product identity.

Where InteliG Fits

The Signal Method is the playbook. InteliG is the engine.

On its own, the Signal Method gives you the structure — the repo layout, the pipeline, the principles. You maintain it manually. You update CLAUDE.md yourself. You push artifacts by hand.

With InteliG, the Signal Method comes alive. Meetings sync automatically into the knowledge layer. Strategy entities push to your repo with one click. Cognis — InteliG’s reasoning engine — synthesizes across all four layers, surfacing insights that connect what was discussed in a meeting to what’s actually happening in the code.

You can use the Signal Method without InteliG. But InteliG turns the methodology from a discipline into a system.

Get Started

The Signal Method repo has everything you need:

Vision and principles — the philosophy behind the methodology
Feature pipeline templates — Mission, Vision, PRD, Design, Architecture, Completion
A worked example — async code review, fully traced from transcript to completion
Agent configurations — prompts for AI reasoning, classification, risk detection
A quickstart guide — set up for your org in minutes

github.com/intelig/intelig-signal-method

Fork it. Build something great.

Related: Confluence Is Where Knowledge Goes to Die — Why internal doc tools fail and why the repo is the only knowledge system that matters.

Read more: As a CTO, I Had One Question. Cognis Answered It. — What happens when the Signal Method meets an AI reasoning engine that synthesizes across all four layers.