Recently, researchers have proposed methods such as the Darwin G"odel Machine and the Huxley G"odel Machine, which enable open-ended, recursive self-improvement through self-reference. However, these self-referential self-improvement methods require that the competence required to perform the task coincides or aligns well with the competence required for self-modification. To address this issue, researchers proposed SBCO (Self-supervised Block Coordinate Optimizer), a self-supervised, verifier-grounded harness optimizer. SBCO learns a decomposed bank of verifiers and a harness policy via approximate block coordinate ascent, improving the agent's outputs from its own graded feedback---with a fixed meta-agent and no human labels. The experimental results show that SBCO matches or exceeds a customized self-modifying baseline while using 4-5.5 times less compute budget. Blogger's Review: The SBCO method provides a new perspective for agent self-improvement, achieving efficient improvement through self-supervised learning and verifier-grounded optimization. This method has achieved good experimental results while reducing the computational budget, making it worth further research and application.