Every AI learning product claims to care about outcomes. Most of them measure engagement.
Engagement is easy. You log time-in-app, daily active users, streaks, session depth, return rate. You build dashboards that make a growing number go up and to the right. When you need to raise a round, you have a chart to show. When you need to justify a feature, you run an A/B test and ship whatever gets the number higher.
Growth is hard. Growth is what the learner can do tomorrow that they could not do yesterday, in the real world, without your system in the room. It resists quick measurement. It often moves in the opposite direction of engagement in the short run, because real learning includes struggle, rest, and a slow handover of competence. A learner who genuinely grows uses your product less over time, not more.
At Cognere we have made a choice. We measure our success by capability, not attention. This has concrete consequences that we want to name on day one.
We will not build streaks. Streaks are a commitment device that trades on guilt and loss aversion; they drive metrics without driving learning. We will not engineer daily-login mechanics. We will not ship push notifications designed to bring lapsed users back. We will not optimize the product for time on screen. These are the easy wins every consumer AI app is reaching for, and every one of them is a small betrayal of the person on the other side of the screen.
What we will do is harder. We will work with educators, clinicians, and domain experts to instrument the actual skill our product is supposed to teach. We will measure transfer into unaided performance. We will accept slower revenue growth in exchange for durable change in the learners we serve. We will publish what we learn.
This is not a marketing pose. It shapes what we can build and how fast. It constrains the business model. It makes certain kinds of investor conversations impossible, and certain kinds of learner relationships possible.
We accept the tradeoff. Quietly, and for a long time.