Why I’m Betting Against AI Agents in 2025 (Despite Building Them)

July 21, 2025

AI Native Dev
autonomous agents
LLMs
software engineering

Line graph titled "Error Compounding: Why Multi-Step Agents Fail" showing success rate vs. number of steps. Three lines represent different per-step success rates: 90% (red, steepest decline), 95% (orange), and 99% (green, shallowest decline). A red annotation highlights that at 20 steps and 95% per step, success drops to 36%. The green line notes that even 99% per step only gives 82% success over 20 steps. A dashed red line marks 99.9% at the top.

I’ve built 12+ production AI agent systems across development, DevOps, and data operations. Here’s why the current hype around autonomous agents is mathematically impossible and what actually works in production.

Source: Why I’m Betting Against AI Agents in 2025 (Despite Building Them)

A sobering reflection on the reality of agentic systems, with some thought on how to build them to avoid these fundamental problems.

Related videos