The open-source revenge fork of your corporate project is not usually how you quit your job, but here we are.
The Summary
- Michael Pfaffenberger, creator of Walmart's internal AI coding tool Code Puppy, is leaving for Pydantic after scaling the tool from 1,000 to 75,000 users in a year.
- He simultaneously released an open-source version on GitHub built "angrily in reaction to Windsurf and Cursor removing access to models and raising prices."
- Walmart's version democratized coding across the company, from engineers to hourly store associates. The open-source fork asks if you'd rather "plow a field with one ox or 1024 puppies."
The Signal
Walmart's Code Puppy grew to more than 75,000 internal users across merchandising, supply chain, and store operations. That's not just developers. That's store managers. That's the kind of horizontal adoption that signals something fundamental about where AI coding tools are headed, and it's not toward $40/month seats for professional engineers.
Pfaffenberger, a distinguished engineer in Walmart's Global Tech group, built Code Puppy on the Pydantic AI library. The tool lets anyone write, edit, test, and manage software using natural language. He called it "democratized technology" in his departure memo. Then he turned around and released an open-source version that reads like a middle finger to the AI coding incumbents.
"Who needs an IDE when you have 1024 angry puppies?"
The timing matters. The GitHub version explicitly calls out Windsurf and Cursor for "removing access to models and raising prices." This isn't abstract philosophy about open source. This is a builder who watched the commercial AI coding market consolidate and decided to spike the football into the open.
Here's what the open-source Code Puppy ships with:
- Natural language task understanding and code generation
- Built-in reasoning explanations (similar to Windsurf/Cursor)
- Optional DBOS-backed durable execution that survives crashes
- The ability to run 50 instances simultaneously "if you were insane enough"
The Walmart version will continue running under his former colleagues. But the open-source fork changes the calculation for every company watching their AI coding tool bills climb. You can now deploy the architecture that scaled to 75,000 users at the world's largest retailer, for free, on your own infrastructure.
Pfaffenberger joins Pydantic, the company behind the Python validation library that Code Puppy was built on. That's not a random startup pivot. That's the open-source tooling layer absorbing the people who know how to deploy it at scale.
The Implication
Watch how many companies start deploying Code Puppy internally over the next six months. The wedge isn't "we don't want to pay for Cursor." The wedge is "our non-technical staff need to automate parts of their job, and we can't afford to give 10,000 people $40/month seats."
The agent economy doesn't run on artisanal hand-coded workflows. It runs on tools that regular people can use to build regular automation. Code Puppy just became the proof point that you can do that without Silicon Valley pricing.