skip to content
Meir Dick

The Stack and Process Behind #ShippedWithAI

/ 3 min read

Table of Contents

Part of the #shipped-with-ai series — one build per week, documented honestly.

People ask about the process more than the products. Here’s what’s actually behind the pace.


The base: Laravel + React Starter Kit

Every project starts here. Secure auth defaults, TypeScript, Inertia v2, Tailwind v4. I never solve authentication from scratch — the security surface is too important, the mistakes are too easy to make, and the defaults in this stack are genuinely good.

Starting from the same base every time also means I’m never making foundational decisions under time pressure. The structure is decided. I’m building features, not infrastructure.


Why consistency is the real advantage

This is the part most people miss when they talk about AI-assisted development.

The single biggest unlock isn’t a clever prompt or a new model. It’s giving AI a foundation it already knows cold. When Claude has strong context on your stack — the conventions, the patterns, the way you structure things — it stops guessing and starts contributing. The difference in output quality is significant.

Laravel Boost gives Claude that context for Laravel. Combine it with a consistent base stack, and AI is writing to your conventions from the first message of every session.


Skills: encoding your standards

The second layer is reusable skills. A skill is a structured context file — it encodes knowledge, standards, or processes that you want AI to have available consistently.

This week I shipped two: laravel-prod-ready (a production-readiness eval that runs the same way every time) and growth-product-context (the foundational layer of GrowEngine, open sourced). The goal is the same in both cases: stop making the same decisions repeatedly. Encode them once, run them reliably.

skills.sh/meirdick


The process: before and after every build

Before I start: define the scope, set up the stack, install the relevant skills.

During: build fast, lean on AI heavily for implementation, stay in the loop on architecture decisions.

After: run the prod-readiness eval, surface the gap list, decide what ships now and what goes on the backlog.

The eval is what keeps the honesty in the process. It’s easy to feel done. It’s harder to be done. The skill forces the question.


What this compounds into

One week. Five things shipped or advanced. That’s not heroics — it’s what a consistent foundation makes possible. Same stack, same eval, same process. AI that knows your conventions. Standards encoded once and reused.

The pace isn’t about moving fast and breaking things. It’s about removing the decisions that don’t need to be made twice. That’s the compounding advantage. It gets more pronounced over time.