Why Senior Engineering Skills Are More Relevant Than Ever
Theo - t3.gggo watch the original →
the gist
Traditional engineering skills like automation, linting, and environment configuration are the primary levers for controlling AI agents, making them more valuable now than in the pre-AI era.
The Shift from Manual Coding to Systems Engineering
Modern software development with AI agents requires engineers to pivot from manual coding to building robust infrastructure that steers models. Rather than treating AI as a black box, engineers should treat domain knowledge as code. This involves creating custom lint rules, CI/CD pipelines, and configuration files that encode architectural patterns and project constraints. By moving logic into automated checks, engineers reduce token usage and prevent recurring errors, effectively scaling their own expertise across an entire team.
Steering Agents with Infrastructure
Effective agent orchestration relies on high-quality steering files like claude.md or agents.md. These files should not merely map file locations but should explicitly define project standards, architectural guardrails, and behavioral expectations. Developers should avoid letting agents write these steering files themselves. Instead, developers should observe agent failures during initial prompts, identify the missing context, and manually update the steering files to prevent those specific classes of errors in the future. This iterative process of refining the environment allows non-engineers or junior developers to contribute to complex codebases with higher reliability.
Career Growth and Seniority
Seniority in the age of AI is increasingly defined by the ability to build systems that elevate team productivity rather than individual code output. The most impactful engineers are those who create environments where code can be landed effectively by others. This mirrors the traditional path to staff engineering, where the focus shifts from individual contribution to system design. By building custom tools—such as the author's Cloudflare-based file upload service for PRs—developers can solve specific friction points in their workflow, creating a tighter, more rewarding feedback loop that makes development feel more like high-level systems design.