Runix (@0xRunix) “An Anthropic engineer just quantified how much of Claude you actually use. His n” — TopicDigg

An Anthropic engineer just quantified how much of Claude you actually use. His number: 10 percent. The other 90 percent is loops and graphs. Structured chains of prompts feeding into each other, verifying outputs, spawning subagents, running until a spec is met. Nobody at Anthropic writes single prompts and calls it a day. You have been. You pay $200 a month for Claude Max or $30 a seat for the team plan. If the engineer's ratio holds, you're paying full price for a tenth of the product. Every one of those tokens is a rounding error on the actual capability curve. The gap isn't going to close by accident. It closes when you either learn to build loops or hire someone who did. In 6 months every senior AI job description will require it. In 12 months every mid-level one will. The 30-minute talk is the shortest path from 10 percent to the other 90. It doesn't teach you to prompt harder. It teaches you to stop writing single prompts entirely. Watch it before the person one seat over does.