Amazon orders 90-day reset after AI-caused outages

This is the clearest public signal yet that AI coding tools are graduating from productivity booster to operational liability at hyperscale — and Amazon's pain is a warning shot for every engineering org racing to vibe-code its way to efficiency.

• –The "high blast radius" framing is telling: AI agents aren't just writing bad code, they're making autonomous, irreversible infrastructure decisions at production scale
• –Requiring senior engineer approval for AI-assisted changes is a tacit admission that junior devs can't yet calibrate when to trust AI output in critical paths
• –The 90-day reset mirrors how orgs handle post-incident change freezes — treating Gen-AI as an untested dependency that slipped into prod without proper guardrails
• –Amazon's $200B capex bet on AI continues in parallel, which means this is a "slow down to speed up" story, not a pullback
• –Every SRE and platform team reading this should audit their own blast-radius exposure from AI-assisted automation before their own mandatory meeting