Back to Portfolio
Context Engineering

How Do You Deal With Context Debt?

A system for translating what's in your head onto the page and making it compound over time.

The problem with every new session

Every AI conversation starts blank. Re-explain your project. Re-establish your preferences. Re-describe decisions you already made. With twelve active projects, this adds up fast.

I started calling it context debt: the cumulative cost of reconstructing your working memory from scratch, session after session. The tool is powerful, but it forgets everything the moment you close the window.

What I wanted to solve

Sessions that pick up where the last one left off

No re-explaining. Open a project, and the AI already knows the current status, recent decisions, and what's blocked.

Cross-referencing ideas across domains

An essay on writing style turned into a methodology for how I design deliverables for leadership. A donut chart technique I figured out while building a personal finance tool resurfaced months later when I was building visualizations for a strategic automotive report. The system should surface these connections automatically.

Learning that compounds automatically

Every session should leave the system a little smarter. Lessons get captured, skills get updated, the next session starts with everything the last one learned.

How it works under the hood

Session continuity

A startup hook injects today's date, last commit, and previous session summary. Every session begins oriented with zero manual setup.

Reusable skills and workflows

Executable workflow patterns with routing logic, quality gates, and accumulated techniques. Some load automatically, others trigger on demand. They encode how I work.

Technique almanac

Reusable patterns indexed by problem type and domain. When I start building something new, the system surfaces relevant past experience automatically.

Cross-platform integration

Connected to Google Workspace, GitHub, Supabase, browser automation, and more. The system operates across all of them from one working environment.

Why it compounds

Anyone can clone the structure. On day zero it's empty shells. The value comes from repeated use. Over time, decisions, techniques, and project history build up and become easier to retrieve and apply. After a few months of real work, the system knows your patterns well enough to surface the right context before you think to ask for it.

Try it yourself

I packaged the system so you can try it. Clone the repo, start a session, see how it compounds.

GitHub: Second Brain Starter Kit
Built with Claude Code CLI · Markdown + YAML · Bash hooks · MCP Protocol · Git
Back to Portfolio