When Safety Code Fails: The Hidden Theory Crisis in High-Stakes Engineering

The recent worker fatality at SpaceX's Starbase facility—the latest in a pattern of safety incidents at the site—reveals a deeper crisis in how we think about engineering knowledge. While the tech world debates code quality and programming methodologies, real lives hang in the balance when safety-critical systems fail not just technically, but organizationally.

Peter Naur's 1985 essay "Programming as Theory Building" argued that software development isn't just about writing code—it's about building a living theory of how systems work. This theory exists in engineers' minds, not in documentation. When teams dissolve or rotate, the theory dies, leaving behind brittle artifacts that future maintainers struggle to understand or modify safely.

This framework illuminates why SpaceX's Starbase has higher injury rates than other SpaceX facilities. It's not just about different equipment or processes—it's about fractured institutional knowledge. Rapid scaling, high turnover, and the pressure to move fast means safety theories never fully crystallize in engineering teams. New hires inherit procedures without understanding the underlying mental models that make them work safely.

Consider how this differs from SpaceX's more established facilities. Veteran engineers at Hawthorne have built robust theories around rocket manufacturing safety through years of iteration, near-misses, and careful observation. They understand not just what the procedures say, but why each step matters and how to adapt when conditions change. This tacit knowledge—the theory—is what prevents accidents.

At Starbase, the theory-building process is constantly disrupted. Engineers implement safety protocols they didn't design, following checklists without understanding the failure modes they're meant to prevent. When something goes wrong, they lack the deep mental models needed to recognize emerging risks or improvise safe solutions.

The tragedy isn't just the loss of life—it's that we keep treating safety as a documentation problem rather than a knowledge problem. OSHA investigations will likely find procedural violations, leading to more detailed checklists and training requirements. But checklists can't capture the nuanced understanding that experienced engineers develop about when procedures need to bend and how to bend them safely.

The real solution requires treating safety knowledge as a living theory that must be carefully cultivated in engineering teams. This means slower scaling, longer tenure, and creating space for engineers to build deep understanding rather than just follow instructions. It means recognizing that in high-stakes engineering, the most critical code isn't written in programming languages—it's the mental models engineers carry about how complex systems can fail.

Until we address this theory crisis, we'll keep seeing the same pattern: innovative companies that excel at technical challenges but struggle with the human knowledge systems that keep people safe.