Issues

• Support Google RRSI for regularized harness self-improvement

  Birfy/agentdescent#221 · Filter by author Birfy opened on Oct 10, 2026
  

• 【第 20 个移植】EvoX Genesis(持久递归世界):完全对齐需要三处引擎缺口,以及两处必须声明的偏离

  Birfy/agentdescent#183 · Filter by author Birfy opened on Sep 12, 2026
  1

• Ship AgentDescent as a plugin for DeepSeek Harness, Claude Code, Codex and OpenCode

  Birfy/agentdescent#165 · Filter by author Birfy opened on Sep 5, 2026
  1

• meta_evolve: 演化 evolve() 自己的决策插槽，并在科研 agent 基准上验证迁移

  Birfy/agentdescent#164 · Filter by author Birfy opened on Sep 5, 2026
  2

• 【接 #75】多 lineage：k 个起点并发展开，而不是轮流展开

  Birfy/agentdescent#150 · Filter by author Birfy opened on Aug 11, 2026

• 【6/7 · 算法】把 PPO / GRPO 的决策规则映射到 diff 接受（advantage、自适应 trust region、KL 类比）

  Birfy/agentdescent#76 · Filter by author Birfy opened on Aug 3, 2026

• 【建议】明确 RSI 基础设施定位并扩展并行执行能力

  Birfy/agentdescent#52 · Filter by author zy95-12 opened on Aug 2, 2026
  34