A field guide to vibe coding, told straight
This is the unfiltered diary of Kilnfolk - a glaze & firing journal for studio potters, now live on the App Store. Every prompt that mattered, every bug that humbled me, every dollar it earned. Not a tutorial. A journey, with the tuition baked in.
The uncomfortable truth first
In early 2025, Andrej Karpathy coined “vibe coding”: you describe what you want, the AI writes the code, you steer by feel. The internet heard “you don’t need to think anymore.” That is not what happens when you try to ship something real.
What actually happens is closer to pottery, fittingly. The AI is the wheel - it spins fast and centres the clay for you. But your hands still decide whether you get a bowl or a puddle. The vibe coder’s job is taste, structure, and stubbornness: knowing what to ask for, recognising when the output is subtly wrong, and refusing to accept “it compiles” as “it’s done.”
I’m a product designer who could read Swift but couldn’t fluently write it. In February 2026 I opened a terminal, typed my first prompt into Claude Code, and started building the app my partner - a studio potter - had wished existed for years. This site is the whole story, structured exactly as it unfolded.
The AI writes the code. You own the consequences. That asymmetry is the entire job.
Every number on this site is real, pulled from App Store Connect, Xcode Organizer, and my prompt ledger on 19 July 2026.
The journey, in eight chapters
Why a pottery app and not another todo list. The niche test, 12 interviews in draughty studios, and the moment the idea earned its keep.
Idea → validationThe single highest-leverage move in vibe coding: writing the spec with AI before writing a line of code. Includes my actual two-page PRD.
Spec-first prompting23 iterations of one screen. How to art-direct an AI like a designer, build a token system, and why I abandoned Figma after day two.
Taste as a skillThe daily rhythm that actually ships: small prompts, instant previews, ruthless commits. Plus the 40-minute rule that saved the project twice.
The working loopThree bugs that nearly broke me - a SwiftData predicate crash, CloudKit ghost duplicates, and an iPad keyboard with a mind of its own.
When vibes fail28 potters, 142 pieces of feedback, 12 days. How to recruit strangers who care, and how to triage without losing your mind.
Real humans, real clayWhat you actually need
The toolchain for this entire project cost less than a nice dinner for two, per month. What it required instead was a willingness to read code I didn’t write and kill features I loved.
# prompt-001 · project kickoff ❯ I want to build an iOS app called Kilnfolk: a glaze and firing journal for studio potters. Before any code, help me interrogate the idea. Ask me the 10 hardest questions a sceptical indie App Store reviewer-turned-investor would ask. One at a time. Be brutal. ⏺ Question 1 of 10: There are 40,000 journaling apps. Why does a potter need glaze tracking badly enough to pay for it - and why has every existing pottery app been abandoned within two years of launch? # …and we were off. The full interrogation is in Chapter 1.
The promise
Every chapter mixes narrative with the exact prompts, decisions, and mistakes that shaped it. Nothing is withheld, including the parts where I was wrong - especially those.