Engineering intelligence
Extraordinary claims by coding agents require extraordinary evidence
Factimonious connects repository activity, code changes, tests, and workflows to explain what happened and show the evidence behind the explanation.
Agent summary
Added OAuth callback hardening, and CI now covers the new auth contract.
Constructed sample. The commits, diff, and workflow below are a fixture, not customer source code or a live repository.
Observed
Commit and diff
Branch feat/auth has three commits. The diff edits src/auth/callback.ts and adds a redirect allowlist.
Observed
Tests and workflow
A new contract test, auth.callback.spec, is in the diff. workflow_run #1842 on that branch completed for the contract-test job.
Inferred
Factimonious reading
Callback routes changed, and CI ran the new contract test. The summary is a reading of those artifacts, not a separate fact.
Unavailable · The sample does not show the change is correct or secure, what the agent intended, or any effect on productivity or business impact.
Product
See it in action
The problem
When agents write the code, how do you keep them honest?
AI-assisted change is arriving faster than teams can inspect and communicate it. Commit messages, PR descriptions, and generated summaries can describe work without establishing what the repository contains.
- —Commit messages and PR descriptions increasingly diverge from what actually landed.
- —Standups and retros drift into narrative when they are not tied back to repository artifacts.
- —Teams lose situational awareness as generation speed outpaces human review capacity.
The method
Inputs, analysis, and outputs
Inputs
- Commits and diffs
- Pull requests
- Tests and workflows
Analysis
- Semantic analysis
- Evidence extraction
Outputs
- Standups and retros
- Engineering intelligence
The questions
What changed, why it matters, and what supports it
What changed?
Which commits, diffs, and pull requests actually landed.
Why does it matter?
What those artifacts suggest about the work, stated as a reading rather than a fact.
What evidence supports the conclusion?
The repository signals the reading cites, and the gaps it cannot close.
Learn
Start with a question you already have
These lessons are useful on their own. Each one states what a method can establish and where it stops.
Code Review
Reviewing AI-Generated Code
How to review agent-produced changes when volume is high and PR descriptions may overclaim.
Software Evidence
Evidence vs Assertion
The Factimonious distinction: assertions are claims; evidence is observable support for those claims.
Git Intelligence
Code Churn
What code churn measures in Git history, when it is useful, and why high churn is not automatically “bad engineering.”
AI Engineering Economics
Cost Per Accepted Change
Attribute agent spend to changes that actually merge—so abandoned loops do not look like productivity.
Who it is for
A path for builders and a path for organizations
Individual builders
Connect a repository, read a standup against the diff, and use Learn when you want the method without the product.
Teams and enterprise
Compare repository limits, seats, history, and where governance and retention become configurable.
See the evidence behind the explanation
Connect a repository when you want this on your own work, or keep reading when you want the method first.




