MatrAIx: testing a product is not testing a decision
In August 2026, Harvard and MIT researchers launched MatrAIx: AI-simulated users that test software or a website before release. Kapari does something else: it tests your decision before you announce it.
A price increase has no screen
Software gets tested in users' hands: where they get stuck, you fix it. A price increase or a reorganization plays out elsewhere: who will go along, who will push back, and why. Kapari has your customers, employees and partners react, as simulated voices, and lets you fix things before you announce.
On MatrAIx's launch day, we tested Harvard and MIT's announcement with Kapari. Read the case.
What people ask us
Does Kapari use MatrAIx?
No. The two tools answer different questions.
Do I have to choose between the two?
No. Test your product on simulated users, then test the launch announcement with Kapari.
For the curious: MatrAIx's numbers
The 8.3 billion characters announced are possible combinations: the published base holds about one million profiles. The 91.5% score says each character stays true to its profile in 366 of 400 trials; it measures consistency, not prediction. Paper posted on arXiv on August 4, 2026.