A scrap picker can know their neighborhood better than any pricing dashboard. The information gap is that the other person in a negotiation may be the only source of a price. Scrapwisest is my attempt to give pickers a usable reference before the next sale.
The app lets someone ask for a scrap rate or record a sale in their own language. The quote is scoped to a country, city, and currency: a Karachi rupee estimate should not silently become a rupee estimate somewhere else. Reports are resolved from neighborhood to city to country, with trusted and fresh reports preferred.
Prefer an honest blank over a made-up number
One design rule matters more than a polished answer: if there is no suitable report or seed for a market, the app should say there is no estimate instead of inventing a sale price. The README describes optional public listings for low-confidence estimates, with strict currency and per-kilogram checks. That keeps the product centered on local market context instead of treating exchange rates as a substitute for local pricing.
The everyday workflow has two verbs: ask a rate, or log what was sold. The agent routes those requests, while price math and basic fraud checks stay in code. Language understanding is useful, but it should not be in charge of arithmetic or quietly fill missing data.
See the demo
I recorded a short Scrapwisest demo. The repository has the implementation and setup notes. Scrapwisest is a prototype, not a source of guaranteed or official prices; its estimates depend on available reports and market coverage.
The challenge I wanted to work on is not just “add AI to recycling.” It is how to make a tool useful to someone who needs a local, understandable number at a real moment of negotiation.
Top comments (0)