000 — INSIGHTS

Field notes from the last 20%.

What we learn taking software from almost-done to shipped — and keeping it there. Checklists, decisions and the boring details that decide whether a launch is calm.

01 — Rescue · 10 min read

Your Lovable or Bolt app works in preview but breaks in production: the causes and the fixes

Most AI-built apps that work in preview and fail in production break for the same handful of reasons: environment variables baked in at build time, auth redirect URLs still pointing at the preview, missing database authorization, webhooks that are not idempotent, CORS, and no error monitoring. Each one has a specific, testable fix.

  • 02 — Rescue · 9 min read

    Supabase Row Level Security checklist for AI-built apps (with SQL)

    If your app talks to Supabase from the browser, Row Level Security is your access control. Enable RLS on every exposed table, write one policy per operation using auth.uid(), keep the service_role key on the server, lock down storage buckets, and test as an anonymous user and as a second account. Copy-ready SQL below.

  • 03 — Business · 7 min read

    Rescue, refactor or rewrite? How to decide what to do with an AI-built codebase

    Rescue when the product works and the problems are security, configuration and missing safety nets. Refactor when it works but every change is slow and risky. Rewrite only when the data model or platform is fundamentally wrong — and even then, replace it piece by piece. The deciding factor is the data model, not how messy the code looks.

  • 04 — Engineering · 8 min read

    How to pay down technical debt without stopping feature work

    You pay down technical debt without stopping feature work by treating it as a standing line item, not a project. Measure where debt actually hurts, keep a ranked register, reserve a fixed share of every sprint, and replace risky parts incrementally behind tests, abstractions and feature flags. DORA metrics tell you whether it is working.

  • 05 — Business · 8 min read

    What a credible 99.9% SLA looks like (and why “100% uptime” is a red flag)

    A credible SLA promises 99.9% monthly availability, measured from outside, with clear exclusions, committed response times per priority and service credits when the target is missed. That allows about 43 minutes of downtime a month. A vendor promising 100% uptime or fixed bug-fix times is either not measuring or not telling you the truth.

  • 06 — Compliance · 8 min read

    EU compliance for app makers in 2026: CRA, accessibility, NIS2, GDPR

    In 2026 an EU app maker needs four things in order: the European Accessibility Act has applied since 28 June 2025, Cyber Resilience Act reporting since 11 September 2026, NIS2 is national law in Slovakia and Czechia, and GDPR Article 28 still requires a processing agreement. Most of the work is engineering hygiene you should be doing anyway.