Code 2025
Making Codebases Agent Ready – Eno Reyes, Factory AI
Overview
This talk argues that making codebases "agent-ready" is crucial for unlocking the full potential of AI in software engineering. The core thesis is that robust, automated validation and verification systems within a codebase are the primary enablers for successful and scalable AI agent adoption, rather than solely focusing on the capabilities of the AI tools themselves. By enhancing these internal validation mechanisms, organizations can significantly increase development velocity and the effectiveness of AI-powered workflows.
Who should watch
- AI Engineers
- Product Managers
- Software Developers
- Engineering Leaders
- Anyone building or integrating AI agents into software development workflows
- Organizations looking to improve developer productivity and code quality
Key takeaways
- The frontier of AI capabilities is largely determined by the ability to specify objectives and search for solutions, with software development being a prime example due to its inherent verifiability.
- Tasks that are easier to verify than to solve, especially those with objective truth, scalability, low noise, and continuous signals, are ideal for AI automation.
- Traditional software development often relies on human oversight for tasks with moderate test coverage (50-60%), but this becomes a bottleneck when integrating AI agents.
- Organizations need to move beyond basic linters and unit tests to implement more opinionated validation criteria that can specifically identify and reject "AI slop" while passing high-quality AI-generated code.
- Shifting to specification-driven development, where agents generate solutions based on defined constraints and validation parameters, is key to building reliable AI-powered software.
- Investing in organizational practices and automated validation systems, such as enhanced linters and automated tests, is more impactful than solely comparing the accuracy of different coding tools.
- Well-defined validation loops enable more complex AI workflows, including parallel agent execution and decomposition of large modernization projects.
- The role of software engineers will evolve towards curating the development environment, setting constraints, and building automations that enable AI agents to function effectively.
Notable quotes
*A slop test is better than no test.*
*The limiter is not the capability of the coding agent. The limit is your organization's validation criteria.*
Unofficial community note. Prefer the recording for nuance.