AI Coding meets Code Health 🪄 — with Stuart Caborn

2026-05-29 · Source: Refactoring · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Advanced, extended

Summary

Love Holidays, an online travel agency, has achieved remarkable engineering efficiency by integrating AI coding with a strong focus on code health. Their engineers deploy to production over 80 times monthly, with more than 60% of production code written by AI, while maintaining an elite change failure rate under one percent and perfect code health. This success stems from a deliberate, multi-year journey that prioritized improving code quality for human engineers *before* AI adoption, using tools like CodeScene to identify "hot spots" and make quality visible through data products. They found that higher code quality leads to more successful LLM outcomes and lower token spend, highlighting the mutual benefits of human-centric code practices and AI efficiency.

Key takeaway

For Directors of AI/ML or MLOps Engineers aiming to scale AI coding, prioritize foundational code health and robust data product strategies. Your investment in high-quality, well-documented code directly translates to more efficient and successful AI agent performance, reducing token costs and accelerating deployment frequency. Implement "shift-left" guardrails and integrate AI-driven "skills" for processes and knowledge to empower both human and AI contributors, ensuring quality and fostering a flywheel of continuous improvement.

Key insights

High code quality significantly boosts AI coding success and reduces LLM token spend, creating a virtuous cycle.

Principles

• "What's good for humans is good for AI."
• Prioritize code health before AI adoption.
• Shift left guardrails to prevent waste and rework.

Method

Love Holidays built data products for code health, incidents, and deployments, exposing metadata to AI agents (like Claude) via an MCP server. This enables AIs to access and correlate data. Guardrails, including unit testing and code quality checks, are integrated into the agentic loop.

In practice

• Turn tribal knowledge and SOPs into "skills" for AI agents.
• Use AI to generate and refine project specifications.
• Measure engineering sentiment alongside quantitative metrics.

Topics

• AI Coding
• Code Health
• LLM Efficiency
• Data Products
• DevOps Metrics
• Agentic AI

Best for: CTO, VP of Engineering/Data, Executive, AI Engineer, MLOps Engineer, Director of AI/ML

Related on AIssential

• Vibe coding and agentic engineering are getting closer than I'd like — The increasing reliability of AI coding agents blurs the distinction between casual "vibe coding" and professional "agentic engineering."
• Dario Amodei, hype, AI safety, and the explosion of vibe-coded AI disasters — AI coding tools, while transformative, pose significant risks without expert human oversight and robust safety mechanisms.
• AI Coding Tools: The Productivity Trap Most Companies Miss — AI coding tools accelerate development but introduce significant, hard-to-detect technical debt, especially in legacy systems.
• Thoughts on coding agents — Coding agents fundamentally shift software engineering workflows, demanding explicit context provision and deliberate model-building from developers.
• Cognition’s Scott Wu says AI coding agents shouldn’t replace humans — AI coding agents like Devin aim to augment human programmers by automating "toil," not replace them.
• Harness Engineering: What Every AI Engineer Needs to Know in 2026 — Harness Engineering defines environments for AI agents to reliably generate production code.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Refactoring.