Generative art, simulations, and visual experiments
Splash!
Interaction DNA
User benefit
Test an abstract benefit through direct interaction.
Counter-evidence
If value is unclear before the result, interaction cost can increase abandonment.
CVR Idea Lab
These hypotheses connect play, understanding, and earned rewards to a meaningful next action.
Let people tune color, shape, or motion, then recommend a matching product or plan.
Derived from: Observe × Simulation × Sensory reward
Measure
Simulate a before-and-after state so the product outcome can be experienced directly.
Derived from: Observe × Simulation × Sensory reward
Measure
Add the brand and CTA to the generated visual and make it saveable and shareable.
Derived from: Observe × Simulation × Sensory reward
Measure
Validate each hypothesis against the offer, traffic source, and reward cost. Measure the final conversion—not game completion alone.
Move to a structurally similar expression and recombine the idea.
Hypothesis
High-fidelity physics + direct manipulation produces a satisfying “I did that” effect.
Principles
Expected Metrics
A tour of the development environment for building 203 browser games and interactions. Claude Code-centric workflow, bun + Biome + Vite, and parallel development with Git Worktree.
Bugs encountered during the development of 203 interactive components, tallied from git history. The most frequent patterns and their fixes presented in a Top 10 format.
A practical record of bulk-fixing bugs across 203 interactive components with Claude Code in 6 commits and 100+ files. The entire process of classification, templating, and verification.
A record of continuously updating CLAUDE.md over 3 months across 323 commits. Rules that worked, rules that didn't, and the limits of "advisory" configuration.