Snr AI Solutions Architect, Previously a Microsoft MVP in Teams. Cat Owner, Amateur Fixer-Upper, Rubbish Photographer. Will Write BICEP JSON For Coffee.
Bristol, England
Joined November 2014
- Tweets4.5K
- Following703
- Followers1.6K
- Likes7.4K
Hey @Tesla - could you please take a look at your app? This is often my experience when chatting with support - the chat just ends.
This is pretty cool! Definitely worth looking at if you aren’t already using different models for planning vs implementing
THIS IS A ACTUAL F*CKING GOLDMINE
Anthropic just shared one of the best tricks for Claude Code
The idea: run Sonnet, Haiku 5.5 and Opus 5.5 at once, each doing only what it's best at
Better results 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 fixes
→ Haiku 5.5 swarms the repo in parallel: search, grep, docs, and test audits at 340 tok/s
→ Opus 5.5 stays on call as architect for the big decisions and signs off on diffs
→ you stop burning Opus context on routine tasks and stop flooding 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 git diff)
Sonnet builds, Haiku swarms, Opus reviews
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
> Sonnet 5.5 runs the main session and coordinates code implementation
> search maps the repo on Haiku 5.5 (read-only, 340 tok/s)
> grep finds call sites on Haiku 5.5 (0.05x cost, isolated context)
> audit scans diffs in a temporary worktree (isolation: worktree)
> Opus 5.5 on call as architect, reviews synthesized diff before sign-off
> 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 with Opus 5.5 as architect
2. Create three custom Haiku 5.5 subagents in .claude/agents/: search (maps repo), grep (finds call sites), and audit (runs isolated diff checks in a worktree)
3. Set isolation: worktree on audit so modifications never touch my working branch
4. Add one rule to CLAUDE.md: delegate all repository searches and test runs to Haiku 5.5 subagents to keep main context under 10k tokens
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, it'll come in handy
Craig Chiffers 📞 retweeted
THIS IS A ACTUAL F*CKING GOLDMINE
Anthropic just shared one of the best tricks for Claude Code
The idea: run Sonnet, Haiku 5.5 and Opus 5.5 at once, each doing only what it's best at
Better results 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 fixes
→ Haiku 5.5 swarms the repo in parallel: search, grep, docs, and test audits at 340 tok/s
→ Opus 5.5 stays on call as architect for the big decisions and signs off on diffs
→ you stop burning Opus context on routine tasks and stop flooding 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 git diff)
Sonnet builds, Haiku swarms, Opus reviews
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
> Sonnet 5.5 runs the main session and coordinates code implementation
> search maps the repo on Haiku 5.5 (read-only, 340 tok/s)
> grep finds call sites on Haiku 5.5 (0.05x cost, isolated context)
> audit scans diffs in a temporary worktree (isolation: worktree)
> Opus 5.5 on call as architect, reviews synthesized diff before sign-off
> 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 with Opus 5.5 as architect
2. Create three custom Haiku 5.5 subagents in .claude/agents/: search (maps repo), grep (finds call sites), and audit (runs isolated diff checks in a worktree)
3. Set isolation: worktree on audit so modifications never touch my working branch
4. Add one rule to CLAUDE.md: delegate all repository searches and test runs to Haiku 5.5 subagents to keep main context under 10k tokens
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, it'll come in handy
Hey @praktika_ai - I’ve cancelled my subscription but you keep trying to charge me and your support team closed my ticket.
The hugging face incident is actually fucking WILD.
Definitely with a listen: open.spotify.com/episode/1Dv…
Craig Chiffers 📞 retweeted
Hope your summer is off to a great start, Windows Insiders.
A few weeks ago, we wrapped the first leg of our Windows Insider meetup tour in London.
This leg took us from Seattle to NYC, Hyderabad, Taipei, San Francisco, and London.
What I’ll remember most is the time spent with Insiders in each city: the feedback, the questions, the stories, and the energy you brought to every stop.
Thank you to everyone who joined us. We’ll be back on the road this fall for the next leg. See you soon.
Here’s a look back at London:
Not personally a fan of the newer Logic Apps GUI in Azure. I'd like to be able to click a run on the left pane and be taken straight to it again.
At Microsoft London today attending the D365 Contact Center boot camp lead by @pajruiter
As someone who’s used the feedback program for years, it’s a little frustrating to hear Microsoft are only just taking note of it now for Windows. A lot of these features were baked into the windows system prior to Windows 11 for years.
The team and I have spent the past several months analyzing feedback from the community. What came through was the voice of people who care deeply about Windows and want it to be better. Read this blog post to learn more about what we're doing in response as we look to raise the bar on Windows 11 quality. Please keep the feedback coming, to help us shape the future of Windows together. blogs.windows.com/windows-in…
