A year ago this week, we set up two folding tables in the Launchbox Berks space at Penn State and started sketching the first version of what would become Mimi. This is the story of what happened next.
March: the question
The founding question wasn't "how can AI help older adults?" It was "why does the digital world keep failing them?" We spent the first six weeks doing nothing but interviews — at senior centers, libraries, and three Sunday brunches at the founder's mother-in-law's house.
June: the prototype
By summer we had a working prototype that lived as a browser extension, designed for a single use case: helping someone read a confusing email and decide whether it was safe to click. It wasn't beautiful, but it worked. The first time someone said "oh, that's helpful," we knew we had something.
"Don't build the demo. Build the next conversation." — advisor at Ben Franklin Technology Partners
September: the pivot point
We thought we were building a product for end users. We learned in customer discovery that the people writing checks were organizations — health systems, banks, senior-living operators — and that the most valuable artifact was actually the insight our assistant generated for them, not just the assistant itself.
That reframed everything. Two products, one mission. We rewrote the deck and the roadmap that weekend.
November: Pennovation
The move from Launchbox to Pennovation Works at UPenn gave us proximity to the research community, the hospital systems we wanted to pilot with, and a denser network of operators who'd been through this before. The space matters less than the neighbors.
January: where we are now
Today we are six people, with paid pilots in motion, a private beta running, and a clear next twelve months. We owe a tremendous amount to the partners listed on our home page — Launchbox Berks, Ben Franklin Technology Partners, Pennovation Works, NSF I-Corps, and NDEC — and to every interviewee who answered the same patient questions, again.
Onward.