I've been building browser automation agents with Playwright and the detection cat-and-mouse game is real. The C++ level overrides are interesting but I'm curious how this holds up against Cloudflare's latest checks - they're not just looking at navigator properties anymore, they're timing canvas operations and checking for inconsistencies in the rendering pipeline. Have you tested this against any real-world bot detection beyond the basic fingerprinting libs?
I'm curious how you're handling the self-evolving part in practice. When I built browser automation agents with Claude API last year, the hardest problem wasn't the initial task execution - it was knowing when to stop iterating vs when the agent had legitimately improved its approach. Does autodidact have any safeguards against the agent just spinning in circles rewriting its own prompts, or do you rely on the user to define success criteria upfront?
I ran into this exact problem building browser automation agents that needed to test DB migrations. The real killer wasn't just getting a sandbox quickly, it was that reverting changes after a failed test would take forever with traditional backup/restore. One thing I'm curious about though - how do you handle agents that need to test against production data patterns but can't actually touch real user data? Do you have a synthetic data layer or is that on the user to solve?