Recursive Coding Agents - Raymond Weitekamp, OpenProse
The bottleneck for reliable coding agents is not intelligence but specifying, managing, and verifying their work.
“today's agents are mismanaged geniuses”
Raymond Weitekamp of OpenProse argues that coding agents fail to deliver reliable outcomes not because models lack intelligence, but because of a missing layer for specifying, managing, reusing, and verifying their work, applying recursive language model (RLM) concepts where the context itself becomes the object of computation. This reframes the agent-reliability problem as one of orchestration rather than raw model capability.