No, That's Not a Software Factory — Ryan Cooke, WorkOS
Source
AI Engineer
Author
AI Engineer
Date
Key takeaways · AI-distilled
Cooke says WorkOS's first factory, an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → in a sandboxAn isolated environment where AI-generated code or agent actions run without being able to touch anything real.Full definition → that is prompted and whose PR is merged, produced results hard to distinguish from engineers running Claude Code on their laptops.
WorkOS's agent TARS lives in Slack, Linear and GitHub and uses webhooks to track projects and pick up the next ticket on its own; a PM agent drafts the product doc and splits approved work into tickets.
Cooke calls an internal MCPThe Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.Full definition → gateway that connects every tool, Snowflake included, the single best investment for a team starting a software factory.
WorkOS measures the factory by outcomes, defect rates, recovery time and adoption, rather than PR counts or lines of AI-written code.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
Why it matters
Offers a concrete counterpoint to hype: measuring a software factory by defect rate and recovery time rather than PR count, plus a working architecture built around a single MCP gateway connecting every internal tool.
