The article outlines how companies are navigating the integration of AI into coding workflows, comparing it to the levels of driving automation established by the NHTSA. The author identifies five levels of automation in coding, starting with Level 0, where manual coding is the norm. At this stage, tools like AI are minimally utilized, serving mainly as enhanced search engines. As teams progress to Level 1, they begin to delegate specific tasks to AI, like writing unit tests or adding documentation. This increases speed but doesn’t fundamentally change the coder’s role. Level 2 introduces a more collaborative experience, where developers work alongside AI, significantly boosting productivity. However, there's a risk of complacency here, as many developers feel they’ve reached their peak capability. In Level 3, the dynamic shifts again. Developers transition from coding to a managerial role, overseeing AI-generated code. This can lead to feelings of frustration as they find themselves buried in reviewing and managing rather than creating. Level 4 marks a further shift, where the developer's role becomes that of a project manager, primarily focused on specs, planning, and oversight. Finally, at Level 5, the process becomes so automated that humans are largely removed from the creation of software, resembling a "dark factory" model where AI handles everything. The author notes that only a small number of teams are experimenting with this level of automation, hinting at a potentially significant future for software development.