22 Dec 2025 Search

previously: beating the high price of price discrimination, personal AI in the rugpull economy

In today’s news: Instacart scraps AI pricing tests that made some products more expensive.

Instacart is ending its AI-powered pricing tests, which led some users to see higher prices on certain products than others….Instacart says it will no longer allow retailers to use the AI-powered Eversight technology it acquired in 2022 to run pricing tests, though retailers may choose to vary the price of items on a store-by-store basis.

Several NGOs worked together on the research that exposed this practice.

In the medium term, though, the decision to do surveillance pricing was a big loss if you consider the opportunity costs. Surveillance pricing services and the internal teams with surveillance pricing skills are costly—and have the skills to increase their own price to capture the anticipated value of the surveillance pricing scheme on average.

And it’s hard to predict how well shoppers and regulators will be able to limit how effective surveillance pricing is. If Instacart bought Eversight with the expectation of surveillance pricing revenue, then Eversight’s investors sold at the top and Instacart will take a loss. So are they going to wait for things to cool down and then bring it back? And Instacart was only caught doing surveillance pricing because people at several different organizations, working together, did a research project on them. They managed to coordinate 400 volunteers. If the only check on corporate shenanigans is the ability of someone at an NGO to get a complicated project approved, then something like this will be back soon, and is probably already going on at other companies.

Yes, this is yet another tale from the rugpull economy. Dana F. Blankenhorn contemplate The Death of Merchandising in an Online World and points out that Companies that rode their brand names for decades are fading. Store chains that relied on brands in turn are also dying, like Kohl’s $KSS and Target $TGT. Consumers seem to believe there is no enduring value worth paying for.

There is value worth paying for—the problem is that so many smart people are spending so much time figuring out how to snatch value that has already been created away from you, instead of trying to make more. Doc Searls, in When Branding Means Relating, adds, What we really (or at least should) mean by brand today is reputation. How a business obtains that in our still-new Digital Age (now with AI!) is an open question.

The answer to that can’t just be open source software and cooperation by well-intended people. It would be great if everyone thought positive-sum and long-term, but that’s not how real-world business decisions get made. Win-lose games are still a problem in the absence of some legislative and court attention. To put some kind of persistent, independent check on privacy harms like surveillance pricing, we need workable Right to Know for inferences and private right of action. Both of those will act as a tax on creepy stuff and (in the long run) an aid to reputation building. It’s good that we’re seeing stories like this in advance of the 2026 privacy bill season. I hope this will help state legislators focus on the kinds of privacy harms that people care about and where they can make a difference. More: what California got wrong on privacy laws