Back to use cases

Solo Developer

Meet Sam Torres

Sam is a freelance developer working solo on 3-4 client projects simultaneously. They use AI agents heavily — for code generation, documentation, architecture planning. But each project has its own ad hoc structure, and Sam keeps re-explaining the same context to agents every session. There’s no team to create documentation for, so “why bother organizing?”

The Challenge

Solo doesn’t mean simple. Sam juggles multiple codebases, each with its own conventions, tech stack, and client requirements. At the start of every AI session, Sam types a wall of context: “This project uses Next.js with Supabase, the auth flow works like this, the client wants…” The agent produces decent output — until tomorrow, when Sam has to type it all again because the agent doesn’t remember.

The real cost isn’t typing — it’s inconsistency. Monday’s agent session uses one approach to error handling. Wednesday’s session invents a different one because Monday’s context wasn’t available. Sam spends Friday reconciling.

How aDNA Helps

Sam creates a lightweight aDNA vault for each client project. Not the full 14-entity-type treatment — a minimal setup that solves the context problem.

The CLAUDE.md solves 80% of the problem: One file that captures the project’s identity, tech stack, conventions, and current state. Every agent session starts with this context auto-loaded. No more wall-of-text preambles.

STATE.md tracks momentum: Updated at the end of each work session with what was done and what’s next. Monday’s session leaves breadcrumbs for Wednesday’s session.

Decisions don’t get lost: Sam’s what/decisions/ directory captures “why this tech stack?” and “why this auth approach?” — questions that come up months later when the client asks for changes.

What Their Vault Looks Like

client_project.aDNA/
├── CLAUDE.md            # Tech stack, conventions, client requirements
├── STATE.md             # Current sprint focus, last session notes
├── what/
│   ├── context/         # 2-3 files: architecture, API conventions
│   └── decisions/       # Key technical choices with rationale
├── how/
│   ├── sessions/        # Lightweight session logs
│   ├── backlog/         # Ideas and client requests
│   └── templates/       # PR template, deployment checklist
└── who/
    └── governance/      # Client communication preferences

Total overhead: 10-15 files. Time to set up: 30 minutes with the base template.

Outcome

Agent sessions start productive instead of starting with context-setting. Consistency improves because every session loads the same conventions. Sam can pick up a project after 2 weeks away and the STATE.md tells them exactly where they left off. The 30-minute setup saves hours of repeated context-typing per week.

The tipping point was clear: even a solo developer with a 15-file project benefits from aDNA when AI agents are involved. The governance files aren’t bureaucratic overhead — they’re the context that makes agents useful.