Inspiration
During a late-night "vibe coding" session, I realized AI hallucinations were forcing me into a vicious cycle. I'd prompt Claude or ChatGPT, get bloated responses with unnecessary context, then need to re-prompt to clarify. But here's the problem: I'd run out of tokens or hit rate limits
What it does
Minimy is a Chrome extension that minifies verbose AI prompts in real-time, reducing token count by 60-70% on average while preserving meaning and intent.
How I built it
Tech stack:
- JavaScript (Vanilla)
- HTML/CSS
- GitHub Copilot
- Prompt API ## Challenges I ran into The biggest challenge wasn't building—it was deciding what to build. Through informal interviews, I discovered two distinct user patterns:
Factual/Research prompts (Quora-like): Users ask broad questions and need comprehensive answers. Coding prompts (Iterative debugging): Users prompt → get response → prompt again → iterate. Here, minification is crucial because each round-trip is a token sink and hallucination risk.
Accomplishments that I am proud of
- 60-70% average token reduction across real-world prompts (verified with examples: code reviews, creative writing, technical docs)
- Privacy-first architecture
What's next
- Live tracking of tokens saved and carbon footprint
- Persona based options for different LLMs as people tend to reserve specific uses for specific LLMs