Project Overview
This recipe scaler was built as a one-shot project - conceived, designed, and implemented in a single pass using AI-assisted development. The goal: create a practical tool that solves the universal problem of scaling recipes up or down without doing mental math while hungry.
The "One-Shot" Philosophy
Unlike the previous Good Vibes Only projects which used iterative feedback cycles, this project was built in a single creative burst. The AI agent was given:
- A brainstorm document with 8 potential project ideas
- The selected concept: Recipe Scaler
- Reference to existing project structure (001, 002, 003)
- Instructions to "have fun" and ship it
Why One-Shot?
Sometimes the best approach is to trust the AI's judgment and let it run. The recipe scaler is a well-understood problem space with clear inputs and outputs. No API integration, no backend, no complex business logic - just pure frontend fun.
This approach works best when:
- The problem domain is well-defined
- No external dependencies or APIs
- The feature set is bounded and clear
- It's a tool you've used before (so you know what's important)
AI Agent Used
Built using Claude Opus 4.5 via Clawdbot, an AI gateway that manages conversations across Discord, Telegram, and other platforms. The agent (named "Shelldon" 🦀) handled the entire build from brainstorm to deployment.
Technical Architecture
Key Features Implemented
Parsing Challenges
The Regex Gauntlet
Ingredient parsing is surprisingly tricky. The parser handles:
Design Decisions
- Orange gradient theme: Warm, appetizing colors that evoke cooking
- Show original amounts: Trust but verify - users can check the math
- Scale factor display: Makes the multiplication transparent (1.5× feels more intuitive than "serving 6 instead of 4")
- No backend: Privacy-first, works offline, no recipe data leaves your browser
- Textarea input: Paste from anywhere - websites, PDFs, notes apps
What Could Be Added (Future Ideas)
Potential Enhancements
The Brainstorm Process
This project emerged from a brainstorming session that generated 8 potential ideas:
- Recipe Scaler ← Selected!
- Photo Wall Planner
- Practice Timer (for instrument practice)
- Moving Box Inventory
- Plant Watering Tracker
- "What Should We Watch" Randomizer
- Allowance Tracker
- Family Command Center
The Recipe Scaler won because it's universally useful, has a contained scope, and scratches a real itch that everyone who cooks has experienced.
Process Documentation
Open Graph Image
The OG image for social sharing was generated using Google Gemini 2.0 Flash with a prompt describing a warm, kitchen-themed visualization of recipe scaling.
Key Takeaways
- One-shot works for bounded problems: When the scope is clear, trust the AI
- Vanilla JS is underrated: No build step, no dependencies, instant loading
- Fractions are hard: "1 1/2" vs "1/2" vs "½" - parsing natural language is tricky
- Show your work: Displaying original amounts builds user trust
- Fun matters: "Have fun!" as a prompt instruction produces better results
Final Thoughts
The Recipe Scaler demonstrates that not every project needs weeks of planning and dozens of feedback cycles. Sometimes you just need to identify a clear problem, point an AI at it, and get out of the way.
Total time from "let's do the recipe scaler" to deployed application: under 10 minutes. That's the power of AI-assisted development when the problem space is well-understood.
Now go scale some recipes. You're welcome. 🍳
← Back to Recipe Scaler