Detail is an AI engineer that solves technical debt.

Detail writes pull requests that solve gaps in test coverage. Focusing on testing in particular allows us to produce high quality output. We produce pull requests that are typically ready-to-merge, as-is.

Longer term, we will solve other categories of tech debt problems within growth stage codebases.

Product Overview

Detail analyzes the gaps in your test coverage and generates new tests that resolve those coverage gaps.

New tests in Detail should be merge-ready. That means:

• New tests pass as-is.
• New tests add coverage, both objectively (measured at line-level) and subjectively (the tests should cover useful guarantees and make the code area practically safer to modify).
• New tests make idiomatic use of factories, fixtures, and internal testing libraries from the rest of your codebase.
• New tests avoid common testing pitfalls: excessive mocking, side effects, conditional asserts, tight coupling to existing implementation.

Engineers use Detail in two ways:

• Detail Inventory, i.e. Detail generates hundreds of tests for untested areas in the codebase, and engineers retrieve these tests from the Detail UI.

  

  Tests in the Detail Inventory, ready to be added to the securedrop repo

• Suggested Tests, i.e. Detail comments on PRs to suggest new tests that can be added in one click.

  

  Tests generated in response to a new pull requests, which are ready to be added

Typically about 80% of the tests from Detail are added via the inventory (i.e. from the Detail UI) and 20% from PR-level suggestions.

Who we Work With

We typically work with growth stage software companies that are leveling up their software practices. Often they’re undergoing a transition in engineering practices, e.g. the shift from “test if you feel like it” to “most code should have tests” or from “code review upon request” to “all new code should be reviewed.”

Customers deploying Detail at scale are merging about 100 PRs per month from our platform (across teams of ~150 engineers), which equates to about 5,000 lines of test coverage added per month.

Examples include:

• Compliance SaaS company, growth stage, where new code is reasonably well tested but there’s a large backlog of untested code from an era in which the company’s engineering practices were a bit more lax. Using Detail both to save engineers time as they build new features and to rapidly add tests when engineers are doing work in a poorly tested area.
• Application Security startup, growth stage, where the existing test suite of mostly integration tests had become unwieldy (too slow, too flake-prone, hard to modify as new features change behavior). Using Detail to build a new suite with better testing practices and reduce reliance on their existing suite.