Sonnet 4.6 in Claude Code: the agent-first model that turns one terminal into a production engineering team.
Built for AI operators and founders running production stacks — the ones shipping agents, automations, and infra every week and tired of half-finished tutorials. If you already have a venture, this drops straight into your stack. If you're building one, this is the operating system behind the curtain.
One email. Instant access. No sequence, no upsell, no list-trap. I'll only ping you if I ship something that fits.
One email. No follow-up unless I ship something that fits.
Sonnet 4.6 turns Claude Code from a coding assistant into a deployable agent system. Pick the role, write the prompt, and the same model plays architect, builder, reviewer, or debugger — reliably enough to run all four in parallel.
STEP 01Architect reads the codebase, returns a plan. Builder writes the code from the plan. Reviewer diffs the result against the original spec. Debugger runs failing tests, fixes the cause, not the symptom. Same model, four prompts, four behaviors.
# Architect prompt
You are the architect. Read the codebase, return a plan:
1. Files to touch
2. Functions to add or modify
3. Test cases the change must pass
4. Risks and edge cases
Do NOT write code. Plan only.Operator running four ventures across NZ and AU. I ship Claude Code agents, AI infra, and automations in production every week — and write up what actually works so you do not have to burn weekends on it.
Browse the rest of the library, or pin a slot to work with me directly.
AI systems for founders. Built in production, taught in public. Currently scaling DebtOps.ai.