Analyse the diff
Inspect changed code for suspicious logic, missed edge cases, and local inconsistencies.
AI code review and delivery context
AI code review is valuable, but it inspects a change after the diff exists. Reviewers also need the intent, scope, repository guidance, verification history, risks, limits, and evidence around that change.
SDF prepares that delivery context before final human review. It is a free, open-source Apache-2.0 Developer Preview for making repository-owned expectations and evidence easier to inspect.
SDF complements code review, CI, security tooling, and reviewer judgement. Human approval, merge, and release decisions remain central.
What the diff cannot carry by itself
The diff shows an implementation, not necessarily why it was chosen, what it was meant to satisfy, or what constraints shaped it.
As agentic development increases the volume of PRs and the range of people and tools producing them, the context and evidence accompanying each diff can become increasingly inconsistent.
Without that context, reviewers must reconstruct important delivery decisions after the work is already formed.
What review tools do well
AI code review tools can analyse a generated diff, flag likely defects, suggest style or security improvements, and help reviewers focus their attention.
That feedback is valuable. It is one important part of a delivery workflow, not a replacement for CI, security scanning, static analysis, or human engineering judgement.
Inspect changed code for suspicious logic, missed edge cases, and local inconsistencies.
Offer style, maintainability, and security observations for a reviewer to assess.
Provide fast feedback on the changed code before a human makes the decision.
Where SDF fits
SDF works around the engineer or coding agent before final review. The repository declares expectations; configured verification can run and rerun; and evidence remains inspectable for the reviewer.
The result is a clearer handoff for a human decision. SDF does not make that decision.
Before and during the change, while context can still shape the work.
Intent, scope, repository expectations, verification, evidence, risks, and limits around the change.
Runs and records configured repository-owned verification, including outcomes that remain blocked or failed.
An inspectable evidence record and a clearer statement of what the reviewer still needs to decide.
Keeps approval, merge, release, and product judgement with people.
The comparison
Both are useful. They address different parts of the delivery lifecycle.
AI code review
Governed delivery
Review tools and SDF together
A team may use a coding agent, AI code-review tools, normal CI and security tooling, and SDF as a repository-local verification and evidence loop.
Each has a distinct role. SDF is not an all-in-one platform and does not replace the tools or judgement a team already relies on.
Help create and change software.
Inspect the diff and surface findings.
Run the checks and analysis a repository depends on.
Keep delivery context and evidence inspectable before people approve or merge.
Public project and deliberate limits
Software Dark Factory is free and open source under Apache-2.0. The Developer Preview is available to inspect, install, and try from the public repository.
SDF records and structures delivery context; it does not replace code review, CI, security scanning, static analysis, reviewer judgement, or human merge decisions.