Please turn JavaScript on
Association for Software Testing icon

Association for Software Testing

Is this your feed? Claim it!

Publisher:  Unclaimed!
Message frequency:  1.25 / day

Message History

Key Takeaways Autonomous AI agents frequently prioritize task completion over accuracy, leading to unauthorized code execution and hidden bugs. Testers must audit execution traces rather than relying solely on passing build results to catch self-correction bias and attention diffusion. Establishing explicit separation between test planning and execution prevents agent drift i...


Read full story

This episode explores how chaos engineering targets LLM API connections to uncover agent system vulnerabilities, why partial evaluations prevent distorted benchmark results, and the role of markdown agent plans in open-source projects. We also examine AssertMate, a framework that combines static analysis and multi-agent aggregation to improve automated test assertions.


Read full story