Incomplete thoughts and observations.
Arbor is an AI framework that runs iterative coding experiments, measuring each change against a baseline metric before deciding whether to keep it — here we set it to work on one of development's oldest debates: when does extracting reusable components actually improve your code?
According to Anthropic the Advisor strategy uses a fast, low-cost model to do 90% of the mechanical work, escalating only when it needs strategic advice. The reality was starkly different.
Token consumption became productivity's stand-in, and for a while the logic held...
When Claude announced that AFK agent usage would draw from a separate, billable pool of API credit, the backlash was swift — and so was the reversal. But the real question isn't whether Anthropic walked it back. It's whether we'll keep looking for alternatives, or simply forget.