Mason Schroder
Case study 2025–2026

Shipping production code as a PM.

Using Claude Code and Cursor to personally merge 2–3 pull requests per week into the Chameleon codebase — bug fixes, UI improvements, and full features.

Role
Product Lead — writing and shipping code via AI-assisted workflows
Team
Felipe (EM), Engineering team (code review)
Status
Ongoing · 2025–2026
01 Premise

A PM who can ship their own fixes.

At Chameleon, engineering bandwidth is precious — a team of under 15 serving thousands of customers. When I found a bug during spec work or customer research, the old loop was: write it up in Linear, wait for prioritization, wait for a fix. That loop could take weeks for small issues.

AI-assisted development tools — Claude Code and Cursor — changed the economics. I could read the codebase, understand the fix, write it, and open a PR in the same afternoon. The engineers review my code the same way they review anyone's. No special treatment, no shortcuts.

Claude Code + Chameleon Dashboard
Claude Code terminal alongside the Chameleon dashboard, showing a code change to fix badge labels on tour analytics
AReal code, real PRs. Every change goes through the same review process as any engineer's work — tests, QA, code review.
BWhy it matters. A PM who can fix their own bugs shortens the feedback loop from weeks to hours. Engineering stays focused on architecture and features.
02 Volume

2–3 PRs per week, personally.

Terminal — building CORE-58 minimize feature
Terminal output showing a feature build for CORE-58, the minimize experience, with Copilot tool support for Minimize Experiences

Over the course of a year, I merged roughly 100+ pull requests into the Chameleon codebase. These ranged from single-line copy fixes to multi-file UI improvements. The screenshots show the two sides of the workflow: Claude Code in the terminal building alongside the Chameleon dashboard open for verification.

The key insight wasn't that AI writes perfect code — it's that AI makes the reading fast enough that a PM with design and HTML/CSS background can navigate a React codebase productively. I'm not replacing engineers. I'm handling the long tail of small fixes that would otherwise sit in the backlog.

03 What I shipped

Types of changes.

  • Analytics label fixes — updating badge text from "users" to "views" to match the actual event being counted.
  • Theme consistency — ensuring in-product experiences respect the customer's configured theme colors.
  • UI polish — hover states, tooltip copy, spacing adjustments across the dashboard.
  • Quick Wins program — 100+ small improvements shipped across the year alongside larger feature work.
04 Reflection

The product thinking is the hard part.

The tools don't replace product judgment. They amplify it. Knowing what to fix and why is still the PM's job — the AI just makes the execution accessible. The most valuable PRs weren't the most complex ones; they were the ones where I caught something during customer research that would have sat in a backlog for months.