Articles on code craft, architecture, and the human judgment that separates working software from software that lasts.
Thoughts on how an old philosophical debate resurfaces in the agentic coding discussion. Is code really disposable or is it an essential artifact for thinking? How does the relationship we have to the coded representations shape our decision-making process?
Real euro figures for an EU-only Rails stack: Hetzner, Scaleway, and OVH hosting, Mollie payments, EU-region email and monitoring, plus where your customer data actually lives. MVP through growing, with the Schrems II and Google Fonts angle no dollar breakdown covers.
The mistakes founders make building software with AI cluster around four missing guardrails: the boundaries AI quietly crosses, code you cannot read, the happy path mistaken for the feature, and a stack that gets rewritten three times. Real failures, real quotes, and the rule to put in place for each.
The names are the system's nouns. 70% of the Rails code I ship is AI-written; the 30% I keep is the part that names the system. Trust zones, the defensive-code creep that follows when naming collapses, and why architecture matters more than ever now that agents generate code faster than anyone can read it.
Real dollar figures for hosting, object storage, Stripe fees, email, monitoring, and backups across Fly.io, Kamal+Hetzner, Hatchbox, and Render. MVP through validated and growing.
What working with me looks like when we build features in a Rails app with AI: spec, tests as a readable definition of done, "go" lock gate, implement, verify. The development loop for non-technical founders.
A 90-minute guide from a fresh Mac to a production-grade Rails app, generated by Claude Code. No development experience required.
ThoughtWorks identified a new category of engineering work between coding and delivery. Nobody has named it yet. I've been doing it.
Ruby is the fastest, cheapest, and most token-efficient language for AI coding agents. Here's the data.
Constraints as understanding — why working within limits produces the deepest insight.
Signal over noise — why the domain model matters more than the library layer.
Strategic sacrifice — accepting complexity now for positional advantage later.
Embodied mastery vs. pattern-matching — AI as collaborator, not replacement.
Technical debt as physical decay — and the energy required to reverse it.
Why every codebase drifts toward disorder — and what to do about it.
Trained perception — what reduced listening teaches you about reading code.