AI code review and delivery context

Is AI code review enough?

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

A code diff is not the whole delivery story.

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.

Original intent and acceptance criteria Repository-owned standards and guidance applied Scope, limits, risks, and unknowns Verification history, including failures and blocked runs Evidence links and the human decision boundary

What review tools do well

Useful feedback on changed code.

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.

Analyse the diff

Inspect changed code for suspicious logic, missed edge cases, and local inconsistencies.

Surface useful findings

Offer style, maintainability, and security observations for a reviewer to assess.

Help reviewers focus

Provide fast feedback on the changed code before a human makes the decision.

Where SDF fits

Context before final review.

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.

When it starts

Before and during the change, while context can still shape the work.

Primary focus

Intent, scope, repository expectations, verification, evidence, risks, and limits around the change.

Verification

Runs and records configured repository-owned verification, including outcomes that remain blocked or failed.

Reviewer handoff

An inspectable evidence record and a clearer statement of what the reviewer still needs to decide.

Human control

Keeps approval, merge, release, and product judgement with people.

The comparison

Review feedback and delivery context.

Both are useful. They address different parts of the delivery lifecycle.

AI code review

Useful feedback on the diff

When it starts
After a diff has been generated.
Primary focus
Findings and feedback in changed code.
Verification
Can help reviewers consider issues in the diff.
Reviewer handoff
Comments and observations attached to the review.
Human control
Informs a human review decision.

Governed delivery

Context and evidence around the change

When it starts
Before and during the change, while context can still shape the work.
Primary focus
Intent, scope, repository expectations, verification, evidence, risks, and limits around the change.
Verification
Runs and records configured repository-owned verification, including outcomes that remain blocked or failed.
Reviewer handoff
An inspectable evidence record and a clearer statement of what the reviewer still needs to decide.
Human control
Keeps approval, merge, release, and product judgement with people.

Review tools and SDF together

A complementary delivery loop.

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.

Coding agents

Help create and change software.

Review and analysis tools

Inspect the diff and surface findings.

CI and security tooling

Run the checks and analysis a repository depends on.

SDF and human review

Keep delivery context and evidence inspectable before people approve or merge.

Public project and deliberate limits

Inspect the Developer Preview.

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.