Category: Development
-

How to Talk to AI Agents: Closing the Communication Gap, a Beginner’s Q&A
Most disappointing AI results are not a model problem. They come from the gap between what is in your head and what you actually wrote down. A plain-English Q&A for people just starting with AI agents: why “I want this” is never enough, the six things the agent does not know about you, how much…
-

Claude Code for Non-Developers: How to Keep Building Your App After the Prototype Without Breaking It
Moving from Lovable or Bolt to Claude Code gives you real power over a real repo — and real ways to break things. The three guardrails to set up before your first prompt (CLAUDE.md, a test command, branches), the prompts that work for founders, and what to never let it do unsupervised.
-

Vibe-Coded App Backups: You Don’t Have One Until You’ve Restored It
“Do you have backups?” has a trick in it. The real question is when you last restored one. What a backup has to mean for an AI-built app — automatic, point-in-time, off-site, files and config included — the quarterly restore drill, and what to do right now if the AI already dropped the table.
-

7 Schema Smells in AI-Generated Apps (and Which Ones Actually Mean “Rebuild”)
The check that decides harden-or-rebuild, written out. Seven data-model smells we find in vibe-coded apps — JSON blobs, missing owners, bolted-on multi-tenancy, string roles, float money, boolean status, no history — what each costs later, and the only one that justifies re-spining.
-

Pricing a Subscription App for Global Markets: Purchasing Power Parity and What RevenueCat Handles
Flat global pricing leaves money on the table in strong-currency markets and prices you out of weaker ones. What the stores handle automatically, what purchasing power parity actually means in practice, and where the decision is still yours.
-

Stripe in a Vibe-Coded App: Webhooks, Idempotency and the Test-Mode Trap
The first test payment works, which hides that the parts keeping money and access in sync were never built. Five things to check on every AI-built Stripe integration — redirect vs webhook, signature verification, idempotency, mixed test/live keys, entitlements — and the one test that finds most of them.
-

How to Add Tests to a Vibe-Coded App That Has None: The Five Flows That Matter First
Every rescue guide says “get a test suite.” This is the how, for a founder who’s never written one: pick the five flows that would end your week if they broke, write them as Playwright end-to-end tests, wire them to run on every change — a day’s work that catches most “the AI broke something…
-

How to Integrate RevenueCat in a Flutter App: iOS + Android Subscriptions from One Codebase
Setting up in-app subscriptions twice — once for App Store, once for Play Billing — is where most Flutter subscription apps lose a week. How we wire up RevenueCat so iOS and Android share one entitlement layer.
-

Authentication in Vibe-Coded Apps: 8 Holes Around the Login That Works
Login is the part of an AI-built app that always works. The holes are around it — unverified accounts, reusable reset links, sessions that survive a password change, no rate limits, client-editable roles, duplicate accounts, deletion that doesn’t delete. Eight two-minute checks and the fix for each.
-

The Vibe-Coded App Handoff Checklist: 15 Things to Give an Engineer So the Fix Starts on Day One
Handing an AI-built app to an engineer? “Here’s the login” costs you three days. The 15-item handoff package — code, how to run it, service access, what it should do, where it hurts — that turns day one into work instead of archaeology.