Agentic Coding Magnifies Your Infrastructure Problems

July 14, 2025

Agentic coding tools expose gaps in your development infrastructure.

Local development

If you have local development issues, agentic coding tools can accomplish less on your team's machines.

Have to reset your docker containers all the time? Have to reset the database schema? That means more manual intervention to get your environment in the right state for developing new features.

The more bespoke config and setup time you have also makes it harder to use cloud-based coding agents like Devin and OpenAI Codex.

CI / CD

If you have CI / CD issues, agentic coding tools have to do more work to get code changes into a state that's ready for PR.

Have a flaky test that needs to be rerun every 5th run? Have CI steps that take 30 minutes? Only release once a day? These are all gum in the works for agentic coding tools.

Deployment complexity

The more services, data stores, SaaS platforms, and external dependencies you're orchestrating increases the risk profile of changes from agentic coders because modeling your system becomes so much more difficult.

Logging / Observability

Have 5 logging and observability platforms? It's harder for agentic coders to guide investigations because the breadcrumbs are scattered around different tools.

Bottom line

Invest in your development infrastructure. This will pay enormous dividends over the next couple years as agentic coders get more and more powerful.