I use AI as a practical creative and technical collaborator: to clarify requirements, turn ideas into sensible phases, speed up implementation passes and reduce the friction around documentation.
The goal is not to hand over ownership. The useful version of AI makes work easier to review, easier to ship and easier to understand later.
Goals
- Move from ideas to shippable increments faster.
- Keep implementation reviewable and maintainable.
- Reduce repetitive delivery admin.
- Capture decisions as work progresses.
- Use AI to support judgement rather than replace it.
Approach
Begin with a small brief
Start with a concise description of what good looks like, list the hard constraints and define a clear acceptance checklist.
Use ChatGPT for structure
Turn the brief into phases, identify risks early and shape prompts that are precise enough for implementation work.
Use Codex for targeted passes
Apply changes in small slices, reuse existing repo patterns and verify each pass with builds, type checks and practical sanity checks.
Use voice tools to lower writing friction
Talk through ideas first when that is easier than staring at a blank page, then tighten the result into useful documentation or a case-study narrative.
Keep local experimentation local
Use local models for generation and validation where that makes sense, including future game-content and playability experiments, without connecting model output directly to live products.
Results
- More consistent momentum from brief to implementation.
- Smaller, more reviewable changes.
- Better documentation as a by-product of the work.
- Less repetitive effort without giving up manual quality control.
- A workflow that stays useful because it supports existing habits rather than demanding a shiny new system.