Agents

Introduction: Agents Are at Their DOS Moment In 2025, AI agents are exploding in capability. Tools like Claude Code can write code, run tests, fix bugs, and autonomously complete complex engineering tasks. For many people, this feels like the second major shift since ChatGPT first appeared. But if you look closely at how today’s agents actually operate, a quieter and more uncomfortable truth emerges: Their foundations are still extremely primitive. Most agents today manipulate your file system and terminal directly. There may be confirmation prompts or guardrails, but the underlying model remains trust-based, not isolation-based. Safety depends largely on the agent behaving well. This should feel familiar. It closely resembles DOS-era computing in the 1980s. DOS worked. You could write programs, edit files, and build real software. But it lacked nearly everything we now associate with a modern operating system: No memory protection No true multitasking No standardized device abstraction Applications talked directly to hardware. Developers were responsible for everything. AI agents are standing at the same starting line today. What took traditional computing nearly three decades—from DOS to Unix, Windows, and modern kernels—will likely replay in a much shorter window for agents. ...

Most people start with automation because they want to save time. They end up discovering something more important: automation saves mental energy. Automation Reduces Actions. Agents Reduce Decisions. Traditional automation follows rules. AI agents handle situations where rules break down. The difference matters. Automation removes steps. Agents remove repeated thinking. Why AI Agents Feel Powerful (and Dangerous) Agents feel powerful because they: Interpret context Decide what to do next Act without waiting for permission They also fail quietly. When an agent makes a wrong decision, it often looks “reasonable” until consequences appear later. The Hidden Cost of Over-Automation Many agent projects fail not because they don’t work, but because: No one is accountable Outputs are trusted too early Edge cases are ignored A useful agent keeps humans in the loop—especially at the boundaries. When You Actually Need an Agent You likely need an agent only if: The task repeats frequently The decision criteria are stable Errors are reversible If mistakes are expensive or public, slow down. The goal isn’t autonomy. The goal is less cognitive load with controlled risk.