Jev 承接 Claude Code 的快速决策层

elune @elune0x原推发布于 2026.10.04

在 X 上打开原推

Claude Code tip: once Opus 5.5 is your main model, don't let Fable 5.1 sit unused. Keep it on call through /advisor. Run /advisor fable. Opus 5.5 keeps writing the code. Fable 5.1 sees the full session, including every tool call, and weighs in at three points: → before a plan: is this the right approach? → when the same error returns: am I looking in the wrong place? → before saying "done": what did I miss? Fable 5.1 reviews. Opus 5.5 ships. Jev engineering uses the same approach one layer down: quick decisions (which file, which tool, retry or stop) go to Jev in under half a second. The big model only gets the decisions that call for deeper thinking. - the full tree > Opus 5.5 on high runs the main session > explorer reads the code > worker edits and runs tests > researcher looks up the docs > all three on medium > Fable 5.1 stays on call as the advisor Paste this tree and prompt into Claude Code ↓ "Rebuild my Claude Code setup around this tree: 1. Check ~/.claude/agents and .claude/agents for subagents that already fit the explorer, worker and researcher roles. > Draft new ones only for roles that are missing > Set each to model: opus and effort: medium > Skip subagents that pin a different model and list them 2. Set the main session to high using effortLevel in ~/.claude/settings.json, and set advisorModel to fable 3. Check for anything that could disable the advisor (CLAUDE_CODE_DISABLE_ADVISOR_TOOL, DISABLE_TELEMETRY, any variable that prevents feature-flag fetching), plus CLAUDE_CODE_EFFORT_LEVEL, which overrides subagent effort. Report what you find without changing it 4. Add one rule to ~/.claude/CLAUDE.md: consult the advisor before a large plan, when an error repeats, and before calling a long task done Show me every proposed change as a diff first. Don't edit anything until I say go." ↳ http://code.claude.com/docs/en/advisor

这条是引用着另一条写的：x.com/0xwhrrari/status/2106731085346304496