Skip to content

Product Vision: ChatKcal

What is this?

The core philosophy and roadmap for ChatKcal, defining the "Anti-Database" model and the "Sidecar" approach to AI-driven nutrition tracking.

Core Philosophy: The Anti-Database

(Existing content...)

The Guest-First Onboarding

We believe users should experience the "Aha!" moment (logging their first meal) before being asked for an email address.

  • Anonymous Entry: Land and log immediately.
  • Local Sovereignty: Data belongs to your device (IndexedDB) until you choose to sync.
  • Frictionless Promotion: One-click account creation seamlessly migrates your local history to the cloud.

The "Sidecar" Model

ChatKcal does not try to rebuild the AI. Users already pay for ChatGPT Plus, Gemini Advanced, or Claude Pro. They have access to the world's best vision, reasoning, and context models.

The Problem: AI models have "amnesia." They are great at analysis but terrible at long-term data storage and trend tracking. The Solution: ChatKcal is the long-term memory for your AI.

  • You talk to your AI (ChatGPT/Gemini) to analyze the food.
  • You deposit the data into ChatKcal for storage, visualization, and progress tracking.

The Roadmap

"The End of the Search Bar" We are removing the friction of the copy-paste workflow.

  • Mechanism: A "System Prompt" installed in the user's AI instructions.

  • Workflow:

    1. User sends photo/text to AI.
    2. AI generates a deep link (e.g., chatkcal.com/log?meal=...).
    3. User clicks the link.
    4. Meal is logged instantly.

Phase 2: The Prompt Library

To facilitate the "Bring Your Own Intelligence" (BYOI) model, we will provide a library of optimized system prompts.

  • Prompts for "Strict Keto" analysis.
  • Prompts for "Loose Estimation" (speed over precision).
  • Prompts for specific cuisines (e.g., "Singaporean Hawker Expert").

Phase 3: Smart Paste & Formatting

Making the app resilient to "messy" inputs.

  • Users should be able to paste a raw paragraph, a JSON block, or a Markdown table.
  • Our internal parser (or a lightweight local LLM) cleans it up before storage.

Target Audience

  • The "Pro-sumer": Tech-savvy, likely has a ChatGPT/Claude subscription.
  • The Developer: Appreciates JSON, data sovereignty, and "no-nonsense" tools.
  • The Frustrated Tracker: Someone who has quit MyFitnessPal 5 times because manual logging was too slow.

Design Aesthetic

"Linear-style" / "Radix"

  • Minimalist.
  • High contrast (Black/White/Zinc).
  • Sans-serif typography.
  • Data-dense but not cluttered.
  • "Tools for thought" vibe, not "Health & Wellness" vibe.