Human intent
Make purchases that fit the person's own priorities, including sustainability, health, price and confidence in the available evidence.
← All projects01 / CHOICE / VALUES / 2025-26
Ziran / Green FilterShould I buy this?
Working post-thesis product
A shopping decision system that turns a person's values and constraints into a visible recommendation instead of hiding the reasoning inside a generic assistant.
System / agent action
Extract a structured product object, fill missing context, and return a verdict with supporting facts, alternatives and sources across browser, web and mobile surfaces.
Hard boundary
The AI does not get to silently manufacture certainty. Ziran distinguishes verified, estimated, brand-reported, AI-inferred and unknown information, and the post-thesis product still stops at decision support rather than autonomous purchase execution.
INTENT PRESERVATION / CORE
The AI does not get to silently manufacture certainty. Ziran distinguishes verified, estimated, brand-reported, AI-inferred and unknown information, and the post-thesis product still stops at decision support rather than autonomous purchase execution.
Evidence in the system
- The current product asks “Should I buy this?” and returns a verdict, facts, alternatives and sources.
- Product structure comes before model inference: JSON-LD, Open Graph, DOM and site adapters are used before AI fills gaps.
- The project preserves a clear provenance boundary between the filed MA thesis, evaluated prototypes and later product work.