Commentary
Commentary perspective

Diffs Are Not Documents

Why agent-generated plans, specifications, and documentation need a review surface designed for meaning—not only changed lines.

An argument for using complete artifacts and exact source changes together—while keeping human decisions attached to meaning.

  • 9 arguments
  • 7 minute read
  • Published July 31, 2026
  • By the Commentary team

Agent-generated work changes the economics of writing. A coding agent can create or rewrite a substantial implementation plan before a human reviewer has finished reading the first version. The review interface must respond to that shift.

01

Git diffs are excellent at exact source inspection.

They show additions, deletions, and movement with the precision software teams need. Commentary keeps Raw and Changes available because source evidence matters.

02

A diff is not always the best primary reading interface.

A specification has an argument, sequence, and reader journey. Understanding the complete artifact often needs to happen before inspecting changed lines.

03

Agent-generated prose often rewrites broadly.

An agent may reorganize a rollout plan, normalize terminology, and regenerate a table in one pass. The diff grows while the underlying decision remains conceptually small.

04

Line movement can hide conceptual errors.

A moved paragraph may be visually noisy but semantically unchanged. A quiet sentence can introduce a dangerous permission assumption. Changed-line volume is not decision risk.

Large diffTerminology and section movementHigh source noise · Low decision risk
One sentence“All tenants inherit the new permission.”Low source noise · High decision risk
Changed-line volume and conceptual risk are different signals.
05

Semantic review should happen before implementation.

It is cheaper to challenge the data boundary, user-visible failure state, or rollback trigger while it is still a plan than after an agent has implemented it efficiently.

06

Rendered artifact and diff should complement each other.

Read the plan to understand the system. Use the diff to verify the revision, identify removed safeguards, inspect links, and confirm precise source changes.

07

Feedback must remain attached to selected context.

A comment about rollback ownership belongs beside the rollback section. Stable semantic anchors preserve that meaning even when surrounding lines change.

08

Agents need structured decisions, not loose chat instructions.

Unresolved comments, selected context, blocking state, and revision identifiers let an agent revise the affected area without inventing a summary of the conversation.

09

Final acceptance should remain human-owned.

Automation can draft, retrieve, revise, and verify mechanical conditions. The decision that the plan is ready to implement remains explicit and human.

Example

A rollout plan in Commentary

An agent submits an implementation plan containing a “Rollout and rollback” section. A reviewer selects that section and asks: “Name the rollback owner and define the condition that stops the rollout.”

The agent retrieves the unresolved comment and selected context, revises only the affected section, and returns a new version. The reviewer reads Latest, uses Changes to confirm the precise revision, and marks the decision ready for implementation.

Loose chat handoff

  • Decision is mixed into conversation history.
  • The affected plan section is implicit.
  • Verification depends on memory.

Structured review

  • Feedback stays attached to Rollout and rollback.
  • The agent receives unresolved context.
  • Latest and Changes support verification.

Diffs remain indispensable. They just should not carry every review job alone.

When the artifact is a plan, specification, document, or preview, start with the form people need to understand. Bring the exact source change in when it helps resolve the decision.

Open the plan example
Continue reading

Build a review stage around meaning.

Use the practical plan-review guide, or explore the complete human-agent workflow.