Opinion: AI Coding Agents Need Full Auto Mode to Reach Potential

_xjdr · x · 2026-07-10

A developer shared observations on using the latest models like GPT-5.6 in coding agent scenarios. They argue that without enabling the fully automated `--yolo` mode (executing directly without step-by-step human approval), the agent's value drops significantly, as sub-agents requiring approval are often worse than just running prompts manually. The author admits that while they used to be a staunch advocate of "strictly reviewing every line of code diff," the massive leap in current model capabilities has convinced them that the era of letting agents run in full auto has arrived—a shift they find hard to adapt to.

Related event: Debate Over Fully Autonomous Yolo Mode in AI Coding Agents(3 posts)→

Original post →

More from coding & agent

coding & agent channel →