Thoughts on mobile engineering, team leadership, and building great products.
Guardrails for Claude Code on my own site repo: a PreToolUse merge hook, 40 bypass tests, a gate that could approve itself, and what each failure taught me.
Over a thousand inline findings from an AI code reviewer, split by priority: what P1 looks like, how I pin a self-review to a commit, and what I never measured.
Fractional CTO in Spain for a startup shipping mobile: who actually needs one, what the first 30 days should produce, when not to hire one, and how it differs from an agency or a full-time CTO.
How to decide between a rewrite, incremental modularization, or leaving a legacy iOS app as it is, from a year-long Objective-C migration that shipped every release on schedule.
An iOS CI pipeline gets slow in the codebase before it gets slow in the config. Why module boundaries come before cache keys, what flaky tests cost, and what a mandatory Xcode bump does to both.
The autopilot pipeline behind this site, and three times a system that reported success was actually failing silently.
Legacy iOS modernization without a feature freeze: split a monolithic app into modules while the team keeps shipping every sprint. What decides module boundaries, and what breaks first.
A practical guide to writing a CLAUDE.md an iOS team actually follows, from someone who runs one across a multi-repo AI-agent pipeline daily.
SwiftUI vs UIKit in 2026 comes down to which framework your team, your timeline, and each screen can afford. A decision framework from shipping both at scale, including a migration whose dual-stack interim ran months longer than planned.
Why I built my own distribution pipeline in versioned code instead of n8n, and the silent-fail bug that came out of it, one a GUI workflow tool would have hidden.
A real llms.txt implementation on a live site, four months in, and the honest answer to whether any AI crawler has ever read it.
AI engineering workflows measured on an enterprise iOS program: where a software engineer agent workflow saved about 30 minutes per engineer per day, where it failed, and the line between agent work and human judgment.
A real Claude Code iOS app development workflow: where the agent is trusted for build, test, and SPM work, where a human stays in the loop, and the honest numbers.
A practical code review workflow built on Claude Code: multi-agent first-pass reviews, what to automate, what to keep human, and the one number I measured after months of daily use.
Notes from putting AI agents into a real enterprise iOS review loop and measuring it, plus the architecture calls behind a decade of shipped apps.
Why 10+ years of experience still matters in the age of AI-assisted development. A practical guide on when to trust AI and when to rely on your judgment.
Questions about engineering leadership, mobile or AI-enabled teams? Send a note.