Programmers become AI babysitters as Codex, Cursor reshape dev workflow
The real bottleneck? Getting humans to explain what they actually want.
A tweet predicting programmers would disappear has aged hilariously—because they've simply evolved into full-time AI babysitters. Developers now spend their days wrangling tools like OpenAI's Codex for code generation, Cursor's aggressive autocomplete, and Runable for automating boring tasks (docs, landing pages). The irony is that while AI handles the mechanical parts of coding, clients still describe features in maddeningly vague terms: 'make it cleaner but more powerful.' The problem isn't the AI—it's getting humans to explain what they actually want.
This reality underscores a deeper truth in software engineering: requirements gathering remains the hardest problem. No AI can parse 'more powerful' into specific API endpoints or 'cleaner' into refactored architecture. The programmer's new role is translator between human ambiguity and machine logic—a skill that's more consultative than technical. As AI tools accelerate output, the bottleneck shifts upstream to human communication, making product-savvy engineers even more valuable.
- Programmers haven't been replaced; they now manage Codex, Cursor, and Runable full-time.
- AI handles boilerplate code like docs and landing pages, freeing time for harder problems.
- The biggest challenge is extracting clear requirements from vague client feedback.
Why It Matters
AI tools shift software engineering from coding to stakeholder communication, making human-centric skills critical.