Tests pass on my 18-core workstation. The server has 40 cores. Spending four hours trying to get the CI to pass a latency test on a 4-core runner. At what point is this "testing theater"?
sorry anthropic, your auto mode classifier is 100% garbage. 0% protections from dumb actions, 100% fail on legit work so far. NO, not one bad thing caught and stops legit work constantly. It's really not "auto" at all.
You tell Claude it screwed up, then it spends 5 minutes drafting a bug report you did not ask for using your subscription tokens. Anthropic, that is unethical.