AnalysisAI AgentsOctober 12, 2026

AWS engineer ran 1,000 agent tasks in a sandbox to find what breaks

Read original source →youtube.com

Kevin Orellana of AWS pointed a coding agent at the company's own sandbox overnight with instructions to break things, then reported which tasks failed. The talk covers failure modes seen when LLMs execute code at scale.

People · Kevin Orellana

1 source

More stories today

Open the live feed