ClaudeCreative is where my PhD research becomes tangible. It’s a Claude Code plugin implementing seven design interventions drawn directly from the failure modes and literacy framework I’ve been developing through my vibe-coding research.
The Seven Interventions
Each intervention targets a specific gap between what AI-assisted development makes easy and what it makes invisible:
- Visible Automation — surfaces what the AI is doing behind the scenes, so you understand the decisions being made on your behalf
- Minimal Diffs — encourages smaller, reviewable changes rather than wholesale code generation, making it possible to understand what changed and why
- Reflection Pauses — introduces deliberate moments to stop and think about whether the current direction is right, before the AI runs ahead
- Conservative Defaults — pushes toward safer, more explicit configuration choices rather than the overconfident defaults AI tends to produce
- Verification Support — provides structured ways to check whether the generated code actually does what you think it does
- Provenance Preservation — maintains a record of what was generated, what was modified, and what decisions led to the current state
- Embedded Literacy — weaves in just-in-time explanations of the technical concepts you need to understand to make good decisions
Why It Matters
The goal isn’t to slow people down or make AI tools harder to use. It’s to make the process transparent rather than opaque. Most vibe-coding tools optimise for speed: describe what you want, get code, move on. ClaudeCreative optimises for understanding: describe what you want, get code, but also get enough context to know whether the code is doing the right thing.
This is the practical output of my Minimum Viable Literacy framework. Not a course, not a curriculum, but scaffolding built directly into the tools people are already using. If vibe-coding is going to be genuinely useful for non-technical creatives, the safety nets need to be in the workflow, not in a separate training session they’ll never attend. The thinking behind these interventions grew from the research direction described in What’s Next: Building Tools That Actually Help.