Legacy code isn't your problem. Your development infrastructure is.

Picture this: A team spends a month debating whether to rewrite a critical service. Architects draw diagrams. Engineers estimate timelines. Managers hedge risks. The debate consumes sprint after sprint.

Here's the alternative: apply hypervelocity principles. Build a closed-loop AI system that can work with the existing code OR rewrite components as needed. The time to do this in an existing codebase varies, but if you're focused I would target a week to get it done. Once you have it, the system makes preserve-vs-replace decisions based on actual velocity metrics, not ivory tower discussions.

This is plowfield development.

Traditional thinking traps us in a false choice: greenfield (start fresh) or brownfield (maintain legacy). But applying hypervelocity principles opens up a third option, plowfield, that moves faster than either.

Plowfield works by giving AI agents the tools to work the soil. The essentials: comprehensive test suites that validate end-to-end behavior, build and deployment processes that run autonomously in minutes, and documented quality criteria that let agents assess independently review code like the humans will. In this environment, autonomous systems actively till the codebase, testing what to keep and what to replace through continuous feedback loops. The AI decides tactically based on what will most quickly check all the boxes, not human debate.

Anti-patterns that kill plowfield:

• Test suites that only cover unit tests, not real system behavior
• Build processes that can't be observed (speed is less important than observability. Once you have observability, then speed matters)
• Quality criteria locked in someone's head instead of in docs the AI can read
• Manual approval gates that break autonomous operation

This is hypervelocity applied to existing codebases. Not avoiding legacy. Making it as malleable as new code.

But it only works if your infrastructure enables it: feedback cycles that tell you whether you broke your product, documented quality criteria, autonomous operation. Without these, you're not plowing—you're still having the debate.

What month-long debates could your team skip with the right infrastructure?