用 Jev 当决策层重构 Hermes Agent 的 harness
转述 TypeSafe 创始人蓝图：按上下文成本路由模型、按需加载工具 schema、动态激活 AGENTS.md、后台跑只读 reviewer，让 Jev 做 Hermes/Codex/Claude Code 的决策层。
Jev Could Rebuild your Hermes Agent’s Harness 🚀🪽 Jev isn’t replacing Hermes, Codex or Claude Code. It could become the decision layer that makes them leaner, cheaper and smarter. TypeSafe founder Diogo Almeida just shared his blueprint: -Rebuild context for each task -Route models by total context cost -Load tool schemas only when needed -Dynamically activate parts of AGENTS.md -Run read-only reviewers in the background -Route sensitive work by security policy 🚀The killer idea: The conversation should not be the agent’s memory. Preserve the evidence. Reconstruct only the context needed for the next decision. 👉This could reshape Hermes compaction, tool discovery, subagents and approvals. Founder’s working document: https://docs.google.com/document/d/1G61uUB0FifUnmmrPzFQojZ3KpczYKmXGpgEXDJ2l_Zg Do your research too and let’s share ?
这条是引用着另一条写的：「Jev 社区项目合集：浏览器 agent 与搜索工具」