Is Learning to Build Software Still Worth It with AI? (My Honest Take)

YouTube · Watch/Read · (gathered 2026-10-14)

• Despite advanced AI agents, learning to build software remains crucial, but the role of engineers shifts from writing code to “intent engineering” and “harness engineering” – effectively communicating desired outcomes to AI.
• The viral “reverse engineer anything” GitHub repo is a misleading example of AI autonomy because the target application is the intent, providing a perfect reference for the agent to validate against, a luxury not present in most development tasks.
• Cole introduces “The Test”: if the next AI model is twice as good, does a skill matter more or less? Skills like prompt engineering matter less, while higher-level planning, validation, and designing comprehensive agent “harnesses” (sub-agents, hooks, workflows) become more critical.
• Engineers must avoid becoming “meat proxies” who simply pass requests to agents without critical thinking; instead, they must actively define intent, build references, and take responsibility for the agent’s output, becoming “engineers of the harness.”