@William__Gitta@chrisalbon Sure, 99% of the time this is not the case. Even if you have tests doesn't mean it handles every case. Just because the tests are all passing doesn't mean the solution is correct.
@dok2001 Loops are actually OP for training ML models. Just tell it you want a x% success rate against the test dataset. It will just keep chugging until you reach it.
Software development doesn’t have as clear of outcomes on the other hand…
@itsalexzajac Yes I think the pendulum will swing back as we learn what’s being created is slop. For the foreseeable future we still need to understand what we ship.
The only thing stopping me from believing that @AnthropicAI is completely delusional is the @bunjavascript rust rewrite. Even then most agentic workflows don’t have a phenomenal language agnostic test suite.