AI makes weak engineers less harmful

AI coding tools like Claude Code have raised the floor for weak engineers by catching obvious errors and producing at least functionally coherent code. The worst engineers now produce output resembling a standard LLM pull request rather than actively harmful work. Some weak engineers have effectively become thin wrappers around AI tools, which is better for their colleagues but bad for their own growth and questionable value for employers. Strong engineers, by contrast, retain enough taste to catch AI errors and don't fall into this pattern. The author predicts companies will soon scrutinize what value engineers add to AI, not just what AI adds to engineers.