Jev runs files and CLI commands for a Claude Code agent team
Opus plans, Sonnet writes code and Fable hunts bugs, while Jev executes files and commands in 16ms without an LLM; the author says this setup built a trading bot that earned $6,000.
I don't get why everyone isn't doing this yet Anthropic dropped a guide that makes Claude Code faster for a fraction of the tokens. The idea: stop burning Opus 5.5 on every line of code. Give the work to a team instead. Architect on Opus 5.5 plans and merges the final PR. Developers on Sonnet 5.5 write the code in separate worktrees. Adversary on Fable 5.1 hunts bugs before every PR. Jev runs files and CLI commands in 16ms, no LLM needed. One command launches the whole team: claude --teammates ux,backend,adversary Then paste this into Claude Code: Configure my Claude Code workspace for autonomous agent teams: 1. In ~/.claude/teams, generate missing roles: set workers to Sonnet (medium effort) in separate git worktree branches, and the adversary to Fable 2. In ~/.claude/settings.json, set CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 and lock the main session to effortLevel: high 3. Initialize tasks.json with file-level locking for direct peer coordination 4. In CLAUDE.md, add a rule requiring an adversary review before opening a PR Show configuration diffs before applying edits. Opus thinks, Sonnet builds, Fable attacks, Jev does the boring part. I used this exact setup to build my trading bot. It's made me $6,000 so far. I didn't write a single line myself, I just reviewed the PRs. Full guide http://code.claude.com/docs/en/advisor
This post quotes another post: x.com/w1nklerr/status/2096964317107077567