The Migration You've Been Putting Off for Years

Episode Summary

Engineering leaders have been sitting on the same migrations for years — monolith to microservices, ORM rewrites, framework upgrades — because the cost of doing them was always 50% of the engineering org for a year. Claude makes those migrations feasible. But not as a single 12,000-line PR. Dan and Dave break down the method that turns a five-year migration into a three-month migration with a skill, a cron job, and a learning loop.

Key Topics

• The anti-pattern: 12,000- to 100,000-line migration PRs. Tempting with Claude, but the absolute wrong way
• The method: pick a small piece, do the first migration with Claude in the loop, then have Claude write a skill from what you learned
• Run the skill on a cron — Claude opens one small migration PR a day while you ship product code
• Why the engineering fundamentals haven't changed: testing infrastructure, shadow reads/writes, upfront planning are still required
• Adding a learning loop so every cron run gets smarter — each PR is less rocky than the last
• Who this is for: engineering leaders, product leaders, and CEOs staring at tech debt they've been avoiding for years

Notable Quotes

• "30,000, 50,000, a hundred thousand line PRs are not the answer."
• "Your three-year migration is now a six-month or a three-month migration. And the most important thing is — you're not spending any time on it."
• "All the core engineering principles that you've already learned are still super, super valuable."

About The Velocity Lab

Dave O'Dell and Dan McAulay work inside engineering organizations every day helping them ship faster with AI. No hype, no BS — just what's working in the field.

Subscribe: RSS