lockstep

Lockstep vs Swimm

Swimm keeps documentation coupled to the code; Lockstep keeps code coupled to the decisions made outside it: in Slack, PRDs, and tickets. They solve adjacent problems: Swimm answers 'how does this code work?', Lockstep answers 'what did we decide, and is this change about to violate it?'.

What Swimm is good at

Swimm keeps code-coupled documentation up to date as the code changes.

Where Lockstep is different

Swimm documents the code; Lockstep captures the product and architecture decisions made across Slack, Jira, Notion, Docs, and PRs, approves them, and enforces them against both engineers and AI coding agents.

Swimm vs Lockstep at a glance

SwimmLockstep
Source of truthThe code and docs coupled to itDecisions made in Slack, PRDs, tickets, and PRs
Question it answersHow does this code work?What did we decide, and does this change violate it?
Staleness modelDocs update when coupled code changesDecisions bind on approval; conflicting PRs get flagged
AI agent roleAgents can read the docsAgents are briefed pre-session and checked at the PR

When to use each

They aren't mutually exclusive. Keep Swimm for code documentation; add Lockstep as the layer that captures the decisions automatically, gets them approved, and enforces them against both engineers and AI coding agents. See the full pipeline.

Frequently asked questions

Do Swimm and Lockstep overlap?
Barely. Swimm is code documentation that stays fresh as code changes. Lockstep is a decision record captured from the tools where decisions actually happen. A team can genuinely want both: Swimm for how-it-works, Lockstep for what-was-decided.
When is Swimm the better choice?
When the pain is onboarding and code comprehension: engineers can't understand existing code, and docs rot. If decisions aren't getting lost and agents aren't shipping against them, Swimm addresses the sharper problem.
Can Lockstep document my codebase?
No, and it doesn't try. Lockstep never stores your source; it records decisions and their source anchors, briefs the people and agents touching affected code, and checks PRs for collisions with what was approved.

More comparisons