A Senior Engineer’s Guide to AI: Knowing When to Build From Scratch

Pedersen’s approach centers on a critical mindset: never trust an output blindly. While he acknowledges that AI models have improved significantly in handling large-scale coding tasks, he maintains that the risk of error remains high for complex assignments. He likens his role to that of a code reviewer, applying the same skepticism to AI-generated drafts that he would to a junior developer’s pull request.

His most stringent rule applies to mission-critical work, such as internal governance policies. When drafting high-stakes documents that require internal defense, Pedersen writes the core language himself. He limits AI to supportive roles, such as generating diagrams or visual artifacts, rather than delegating the substance of the policy. He also warns against using AI for personal communication or sensitive judgment calls involving people, where human values and context are irreplaceable.

For daily tasks, Pedersen advocates for the "two-second rule": if a task takes less time to perform manually than to prompt, troubleshoot, and verify, the automation is a waste of resources. He urges users to actively challenge AI by demanding sources and questioning logic, a practice he finds vital when exploring unfamiliar subjects. Ultimately, he views the technology as a tool for empowerment rather than a replacement for human judgment, noting that the most effective engineers are those who stay suspicious of the machine.