Large Action Models Are Coming For Your Testing Frameworks!
summary
Claims that Claude 3.5 Sonnet and other Large Action Models can operate a desktop and perform agentic tests autonomously, demos show vision‑based clicks and end‑to‑end app use. Toolchain specifics include MCP and Kira, plus Klein, Cursor and specbased workflows used to generate apps and tests; teams report AI code is faster but risky — cited numbers: 322% more vulnerabilities in AI code, 45% exploitable, StackOverflow traffic down 90%, and self‑healing tests work sometimes but are slow and costly.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:09AI’s impact on dev/testing
- 3:40From LLMs to Large Action Models
- 8:13New agentic AI stack
- 10:19AI in software lifecycle
- 16:14Spec-based safer AI coding
- 20:23Quality and speed in AI dev
- 23:12Testing autonomous agents
- 28:09AI tools for faster testing
- 32:21AI test case generation tips
- 35:51Vision-based testing with LAMs
- 37:59MCP streamlines testing
- 39:45Q&A
related talks
30mAutonomous Browsing using Large Action Models
16mTransforming Test Automation with AI and Machine Learning
23m„Complete” In-Sprint Testing Using Collaborative Models
01h 01mRise of the Machines: Can Artificial Intelligence Terminate Manual Testing?
48mSupercharged Testing: AI-Powered Workflows with Playwright + MCP
21mLLM-Powered Unit Testing: Writing Reliable Java Tests with AI Assistance
25mWhat’s Our Job When The Machines Do Testing?
17mAI-Powered Integration Testing