Monkey proofing my release with sonnette 5.5
retained text
So recently I decided to strengthen my deployment pipeline by integrating claude as a qa tester rather than bother my fellow human. I plugged in sonnet 5.5 to my ci post successful deployment to pre-production, to: 1. Check committee diff from pre-production to prod 2. Grab the relevant tickets in linear and the pr descriptions for max co text on each. 3. Attempt to fulfill the happy path + push it to find potential near bugs. My flow is not perfect but I've already caught a number of bugs which I did not catch myself. Runs cost between 0.5c to 1dollar. This works for now but ultimately im unsure how to scale this once I really start dishing out 10 to 20 prs within 24 hours. (Yes my goal is to finish a sprint within a day even if that means slaving claude while I sleep and have 100 success rate). Any one has a similar setup? Any advice? I could optimize by injecting different models but right now thats not what im interested about. Rather it is: 1. How can I make this better? 2. What scalable directions look like or even what are the right questions to ask? Thanks <3
document fetched 2026-10-21 00:40 · canonical https://reddit.com/r/ClaudeCode/comments/1xakqs9/monkey_proofing_my_release_with_sonnette_55