Independent AI Software Studio · New Zealand
I build real software with AI and document what actually works.
First-hand case studies, agent-system architecture, and practical engineering guides from products I build and operate.
Start here
Featured Case Studies & Guides
Engineering guide
Evaluate AI Coding Agents at the Patch, Not the Promise
A practical system for evaluating AI coding-agent changes with task contracts, fail-to-pass tests, pass-to-pass regressions, and diff review.
Engineering guide
Four boundaries to set before handing a repository to an AI agent
A practical guide to scope, context, authority, and verification when using AI agents on a code repository.
Case study
Off the Desk — PapaCoder Field Notes
Part 7 (finale) of Coding Is Conversation. How ADR-0005, draft-only Internal API, WIP=1, and the editorial DAG live in the Labs repo.
Durable knowledge
Engineering Guides
Draw the Map When a Line Isn’t Enough — Graph Engineering
Part 6 of Coding Is Conversation. DAGs, conditional edges, handoffs vs agents-as-tools, and human escalation nodes.
Design How It Stops — Loop Engineering
Part 5 of Coding Is Conversation. Plan→act→observe, WIP=1, max_turns and retry policy so agents stop spinning.
Build Walls Outside the Prompt — Harness Engineering
Part 4 of Coding Is Conversation. Tools, sandboxes, permissions, and evals that enforce policy outside the model—plus Cursor hooks and PapaCoder’s draft-only API.
Quality Comes From Roles — Multi-Agent Engineering 101
Part 3 of Coding Is Conversation. Move past vibe coding with roles, typed handoffs, and gates—plus copy-paste Research/Implement/Review cards.
Ship by Talking — Vibe Coding From First Prompt to Deploy
Part 2 of Coding Is Conversation. Define vibe coding (Karpathy/Fowler), then ship a Today Focus Board with Cursor prompts, AGENTS.md, and Vercel.
From real projects
Latest Field Notes
Secondary archive