AI Pulse · Under the Radar

Someone just released an open-source shopping agent that flat out refuses to take money for ranking products higher. It’s called northcinder, and in three days it picked up 1,198 stars on GitHub, which works out to about 362 a day. That kind of pickup usually means developers are seeing something they actually want, not just something trending.

Here’s what it does differently. It uses a deterministic ranking system, meaning the order it shows you products in follows fixed rules, not an auction where brands pay to move up the list. Before it buys anything on your behalf, it creates a signed purchase mandate, basically a verifiable record of what it was told to buy and under what terms. And it runs fully local on your machine with no telemetry, so nothing gets phoned home to a company server. You can read the details straight from the project on GitHub.

Why this matters if you run a business: the big platforms are building toward a future where AI shopping agents get paid by brands to recommend certain products, the same pay-to-play model that already shapes search results and marketplace listings. This project is a direct rebuttal to that. It’s a signal that a chunk of the developer community wants agentic commerce to work for the buyer, not the highest bidder.

For owners, the practical takeaway is this. If you sell online, don’t assume every AI shopping agent your customers use in the next few years will be neutral. Some will be, some won’t, and knowing which is which will matter for how you price and market. It’s also a reminder that trust and transparency in automated buying decisions are becoming a real differentiator, which is exactly the kind of thing we build into an AI command centre so your own automations stay verifiable and under your control.