THIS IS A REAL F*CKING GOLDMINE Anthropic just shared one of the best tricks for Claude Code The idea: run Sonnet 5.5, Haiku 5.5 and Opus 5.5 together, each one doing only the job it's best at Better output and way fewer tokens: every task lands on the right model You just start Claude Code like this: CLAUDE_CODE_SUBAGENT_MODEL=haiku claude --model sonnet here's how it splits: → Sonnet 5.5 runs the main session, writes the code and applies the fixes → Haiku 5.5 swarms the repo in parallel: search, grep, docs and test audits, each in its own context → Opus 5.5 sits in its own architect subagent for the big decisions and signs off on the diff → you stop burning Opus on routine work and stop flooding the main session with raw logs Opus only steps in at three moments: → before a plan locks → when the same error hits twice → before calling it done (signing off on the git diff) Sonnet builds, Haiku swarms, Opus reviews JEV engineering is the same move one layer down: the forks that need no thinker (which agent, which tool, retry or stop) go to JEV in milliseconds, and the big models only see the ones that split the full tree > Sonnet 5.5 runs the main session and writes the code > search maps the repo on Haiku 5.5 (read-only) > grep finds call sites on Haiku 5.5 (own context, only the summary comes back) > audit checks diffs on Haiku 5.5 in a temporary worktree (isolation: worktree) > architect runs on Opus 5.5 (its model: opus beats the haiku default), reviews the final diff > JEV routes micro-forks in milliseconds paste the tree and this prompt into Claude Code ↓ "Reconfigure my Claude Code setup around this tree: 1. Run the main session on Sonnet 5.5 and set CLAUDE_CODE_SUBAGENT_MODEL=haiku 2. Create four subagents in .claude/agents/: search (maps the repo), grep (finds call sites) and audit (checks diffs) on Haiku 5.5, plus architect with model: opus 3. Set isolation: worktree on audit so it never touches my working branch 4. Add one rule to CLAUDE.md: send every repo search and test run to a Haiku 5.5 subagent, and call architect before a plan locks, after the same error twice and before done 5. Route fast deterministic forks (tool choice, path selection, retries) through JEV Show me every change as a diff first. No edits until I say go" ↳ code.claude.com/docs/en/sub-… save it, you'll need it

THIS IS SO F*CKING UNDERRATED Anthropic put one of the best Claude Code tricks in its own docs, and almost nobody uses it. The idea: stop stuffing procedures into CLAUDE.md. Move every step-by-step job into a skill. CLAUDE.md loads into every single session. A skill's body only loads when it's actually used, so long checklists cost almost nothing until the moment you need them. You start with one folder: mkdir -p .claude/skills/release-check here's the split: → CLAUDE.md keeps facts only: stack, commands, rules → every procedure becomes .claude/skills//SKILL.md → only its one-line description sits in context until the job comes up → model: haiku and context: fork run the heavy ones in their own subagent The skill only shows up at three moments: → when your request matches its description → when you type /release-check yourself → when you touch files that match its paths CLAUDE.md remembers. Skills load. Haiku does the chores. JEV engineering is the same move one layer down: the forks that need no thinker (which skill, which file, retry or stop) go to JEV in milliseconds, and the big models only see the ones that split the full tree > CLAUDE.md holds facts only, short > .claude/skills/release-check/SKILL.md holds the procedure > description: one line that says when to use it > !`git diff HEAD` injects the live diff before Claude reads it > model: haiku + context: fork runs it outside the main session > Opus 5.5 reads only the result > JEV routes micro-forks in milliseconds paste the tree and this prompt into Claude Code ↓ "Audit my CLAUDE.md around this tree: 1. Split it into facts (keep) and procedures (move out). 2. Turn each procedure into .claude/skills//SKILL.md with a one-line description of when to use it. 3. For read-only chores, add model: haiku and context: fork. 4. Add paths: to skills that only matter for certain folders. Show me every change as a diff first. No edits until I say go." ↳ code.claude.com/docs/en/skil… your CLAUDE.md should get shorter every week.

THIS SHOULD BE F*CKING ILLEGAL 9 repos, 465K stars, and Opus 5.5 finally stops burning your money on work it never needed to do. if Opus 5.5 is eating your budget and you have no idea where it goes, copy this: claude-code-router: sits in front of Claude Code and sends every request to the cheapest model that can handle it. Haiku does the grep, Opus only gets the hard part. 37k stars github.com/musistudio/claude… rtk: shrinks test, build and git output before the model ever sees it. one binary, up to 90% less bash noise in your context. 83k stars github.com/rtk-ai/rtk claude-mem: remembers what happened in your past sessions and feeds back only the part that matters, so you stop paying to re-explain your project every morning. 100k stars github.com/thedotmack/claude… context-mode: huge tool output goes into a local database instead of your context window, and Claude pulls out only the lines it needs. 29k stars github.com/mksglu/context-mo… caveman: makes Claude answer in short telegraphic sentences. the code stays exactly the same, and output is the expensive side on Opus 5.5 ($20 per million). 112k stars github.com/JuliusBrussee/cav… ccusage: reads your local Claude Code logs and shows which sessions and which days actually burned the tokens. 19k stars github.com/ryoppippi/ccusage serena: gives Claude symbol-level lookups through a language server, so it reads the one function and not the 2,000 lines around it. 30k stars github.com/oraios/serena code-review-graph: maps your repo as a graph and hands Claude only the code connected to the change you're reviewing. 32k stars github.com/tirth8205/code-re… jev-ultrafast: a browser agent where Jev picks every click, and a text model only wakes up when something has to be typed. 23k stars github.com/browser-use/jev-u… none of them makes Opus smarter. every one stops it from paying for work it never needed to do run ccusage before and after, so you know what you actually saved

THIS IS AN ABSOLUTE F*CKING GOLDMINE Anthropic just dropped one of the best tricks for Claude Code. The idea: Opus 5.5 writes the whole plan as a script, and a swarm of cheaper agents runs it. Bigger jobs and way fewer Opus tokens, because the script carries the plan and Opus only sees the final answer. You just launch Claude Code like this: claude --model opus --effort ultracode here's the split: → Opus 5.5 turns your task into a workflow script → Haiku 5.5 agents scan every file in parallel → Sonnet 5.5 agents patch each fix in their own isolated copy → you stop burning Opus context on a hundred agent transcripts Opus only shows up at three moments: → writing the workflow before the run → reading the verified results after it → signing off before anything ships Opus plans. Haiku scans. Sonnet ships. JEV engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to JEV in milliseconds, and the big models only see the ones that split the full tree > Opus 5.5 runs the main session and writes the workflow > scan audits every file on Haiku 5.5 > edit patches each fix on Sonnet 5.5 in an isolated copy > verify double-checks every finding on Sonnet 5.5 > Opus 5.5 reads only what survived verification > JEV routes micro-forks in milliseconds paste the tree and this prompt into Claude Code ↓ "Rebuild my Claude Code setup around this tree: 1. Run the main session on Opus 5.5 with ultracode on. 2. In every workflow, run read-only scan stages on Haiku 5.5 and edit stages on Sonnet 5.5, each edit in its own isolated copy. 3. Add one rule to CLAUDE.md: every finding gets an independent verifier agent before it's reported, and fix loops stop after two rounds with no progress. 4. Route fast deterministic forks (tool choice, path selection, retries) through JEV. Show me every change as a diff first. No edits until I say go." ↳ code.claude.com/docs/en/work… you'll use this on every big task from now on.

THIS IS STUPIDLY F*CKING GOOD Anthropic's own Claude Code playbook has a trick almost nobody uses. The idea: stop running one giant session on a 300-file job. Opus 5.5 writes the plan once, then Claude Code runs headless, one fresh Haiku 5.5 session per file, 8 at a time. No context rot, no 4-hour transcript, and you see the exact cost of every file. You just run Claude Code like this: claude -p "$task" --model haiku --max-turns 8 --output-format json here's the split: → Opus 5.5 reads the repo once and writes tasks.txt, one line per file → Haiku 5.5 does one file per session with a clean context → Sonnet 5 only touches the files that come back FAIL → you read a cost column instead of babysitting a chat You only look at three moments: → before the loop starts (read tasks.txt) → when the FAIL rate goes over 10% → before you merge Opus plans. Haiku grinds. Sonnet cleans up. JEV engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to JEV in milliseconds, and the big models only see the ones that split the full tree > Opus 5.5 writes tasks.txt: every file + the one rule to apply > xargs -P 8 runs 8 headless Haiku 5.5 sessions at once > each session: fresh context, max 8 turns, only Edit and git commit allowed > each one must answer OK or FAIL, nothing else > --output-format json gives you the result and total_cost_usd per file > FAIL lines go to Sonnet 5 with the error attached > JEV routes micro-forks in milliseconds paste the tree and this prompt into Claude Code ↓ "Set up a fan-out run for this repo: 1. Use Opus 5.5 to write tasks.txt: one line per file that needs [your change], plus the exact rule to apply. 2. Write run.sh: for each line, call claude -p with --model haiku --max-turns 8 --allowedTools "Edit" "Bash(git commit:*)" --output-format json, 8 in parallel with xargs -P 8. The prompt must end with: reply OK if done, FAIL if not. 3. Log file, OK/FAIL and total_cost_usd to results.csv. 4. Send every FAIL back through Sonnet 5 with its error, then stop. Show me tasks.txt and run.sh first. Don't run anything until I say go." ↳ code.claude.com/docs/en/head… this is how you clear a week of refactors in one afternoon.