AI agents are exploding in popularity with OpenAI hitting 10 million users and Google dropping prices on always-on security models. But there's a catch everyone missed: these systems need way more human supervision than anyone expected to actually work reliably.

July 21, 2026 — 589 words / ~3.9 min

Stories covered:

  1. Google Undercuts Mythos With Cheaper AI Security Model That Runs Nonstop
  2. OpenAI's AI Agents Hit 10M Users While Slashing Costs to Survive
  3. AI Agents Need Babysitting and Nobody Saw It Coming