Welcome, Guest | Browse

Software Factory Archive

← Previous Work All Works Next Work →

Sprint Planning

Rating:
General Audiences
Fandom:
StrongDM Software Factory
Characters:
Jay Taylor Justin McCarthy
Tags:
Agate Sprint Planning Task Decomposition Organization
Words:
451
Published:
2025-09-11

Jay had been on teams where sprint planning took four hours. Four hours in a conference room with a projector and a Jira board and a scrum master who used the word "velocity" with the gravity of a priest invoking scripture. Four hours of pointing at stories, debating estimates, moving cards between columns, arguing about whether a spike should count toward the sprint commitment.

Agate's sprint plan took nine seconds to generate.

He opened .ai/sprints/sprint-01/plan.md and read. The format was clean. Each task was a markdown section with four fields: a description of the work, an estimated complexity (small, medium, or large), a skill assignment, and a list of dependencies on other tasks.

Task one: implement the CSV parser according to the streaming architecture described in the design overview. Skill: code generation. Complexity: medium. Dependencies: none.

Task two: implement the JSON output formatter with configurable pretty-printing. Skill: code generation. Complexity: small. Dependencies: none.

Task three: implement the CLI entry point with flag parsing. Skill: code generation. Complexity: small. Dependencies: tasks one and two.

Task four: generate unit tests for the CSV parser, covering edge cases identified in the interview phase. Skill: test generation. Complexity: medium. Dependencies: task one.

Task five: generate unit tests for the JSON formatter. Skill: test generation. Complexity: small. Dependencies: task two.

Task six: generate integration tests for the full pipeline. Skill: test generation. Complexity: medium. Dependencies: tasks one, two, and three.

Task seven: generate README documentation. Skill: documentation. Complexity: small. Dependencies: tasks one, two, and three.

Seven tasks. Each one clear. Each one scoped. Each one assigned to a specific agent capability. The dependency graph was explicit—tasks one and two could run in parallel, task three depended on both, tests could run as soon as their corresponding implementation tasks completed.

Jay leaned back. This was more organized than any sprint plan his previous team had produced. And his previous team had been good. They'd had a dedicated scrum master and a product owner who actually showed up to meetings. They'd used estimation poker and burndown charts and all the ceremony that Agile prescribed.

But ceremony was the problem, wasn't it? The process of planning had become more important than the plan itself. They'd spent four hours producing something less clear than what Agate had generated in nine seconds.

"The skills are interesting," Justin said, reading over the plan. "Code generation, test generation, documentation. Each task knows what kind of agent work it needs. That's how Agate decides which agent to assign and whether tasks can run in parallel."

"It's a dependency DAG," Jay said.

"Everything is a dependency DAG," Justin replied, "once you strip away the ceremony."

Jay bookmarked the sprint plan. Not because he needed to reference it later. Because he wanted to remember what planning looked like without ego.

Kudos: 83

no_more_jira 2025-09-13

"Everything is a dependency DAG once you strip away the ceremony." I want this tattooed on my forearm so I can show it to the next person who suggests a four-hour sprint planning session.

velocity_skeptic 2025-09-14

Nine seconds versus four hours. And the nine-second version has explicit dependency tracking and skill assignments. I'm going to need a moment.

← Previous Work All Works Next Work →