// AI-AUGMENTED WORKFLOW
Senior craft, multiplied.
AI doesn't make you a better developer. Working with AI well does. There's a difference — and it shows up in the quality of questions you ask, the judgment you apply to the answers, and the decisions you make about when to lean on it and when not to. This is how I use it.
The tools behind the tools
ScrapWars runs on hundreds of interconnected items, materials, crafting recipes, and economic relationships. Managing that data manually — in spreadsheets or raw JSON — would burn hours that should go to building. So I built a tool instead.
Scrap Shop is an internal dev website, built with AI assistance: a purpose-built interface for creating, visualising, editing, and managing game data. Items, material tiers, production chains, crafting trees — all editable in a UI designed around the actual data model, not around generic CRUD patterns.
Without AI, I wouldn't have built this at this stage of a solo side project. The tooling would have waited. AI lowered the cost of doing it properly — and that shifted the quality of everything downstream.
"AI made the cost of doing it right low enough to justify — which changed how I work, not just how fast."
Senior judgment, expanded map
Game balancing is a discipline in its own right. So is economic simulation, material tier design, and the psychology of player progression. These aren't my core domain — but they all need to be applied carefully in something as interconnected as a 4X economy.
AI gives me rapid access to the established thinking in those fields: documented approaches, known failure modes, design patterns from shipped titles. I don't treat this as a substitute for expertise. I treat it as a way to compress the knowledge acquisition phase — reaching the point where I can evaluate intelligently, faster.
The senior skill here isn't knowing everything. It's knowing how to interrogate what you find, recognise when a pattern doesn't apply to your context, and make the call. The map gets bigger. The navigation doesn't get outsourced.
"AI expands the map. Judgment is still the navigator."
Real data, real feedback, no delays
Prototypes built on obviously fake data give obviously fake feedback. A UI full of placeholder item names and off-range stats doesn't tell you whether the layout works, whether the information hierarchy makes sense, or whether the data model holds up under realistic load.
I use AI to generate placeholder content that's close enough to real: item names and descriptions that match the tone, stat values in the right range, production chains that are structurally plausible. The content will be replaced — that's the point — but it's representative enough to surface real design problems early.
The same principle applies to layout work with AI-generated placeholder images. They're not shipped. They're scaffolding. The difference matters, and knowing it matters is the point.
"Placeholder content that's representative enough surfaces real problems early — before investment goes in the wrong direction."
The discipline of fast
Not every idea deserves production-quality code. Some things need to be tried before they earn investment — and the ability to spin up a rough implementation, evaluate it honestly, and either discard it or promote it, is an architectural skill in itself.
AI makes the exploration phase faster without making the decision phase less rigorous. A new system mechanic, a UI interaction pattern, a data flow — I can sketch it in a fraction of the time, which means I explore more options before committing. The throwaway code stays throwaway. The decisions are better for it.
This is what AI-augmented velocity actually looks like at senior level: not moving faster toward the first idea, but reaching better ideas before the clock runs out.
"Not moving faster toward the first idea — reaching better ideas before the clock runs out."
// THE TAKEAWAY
The common thread across all of this isn't the tools — it's the judgment about when and how to use them. That's the skill that transfers.