An interview with Rabbit founder Jesse Lyu on the Rabbit R1, criticisms of the AI device, bad reviews, and more; Lyu says ~5,000 people now use the R1 daily
In a frank conversation at the Fast Company Innovation Festival, Ammunition's Robert Brunner and Rabbit's Jesse Lyu unpacked AI hardware's slow start.
Context & Ripple Effects
Rabbit’s R1 entered the market with a deliberately distinct handheld design and a stated aim of avoiding the conventional phone-or-app model, following an early discussion of its design and device strategy. Its initial pitch also depended on external AI capability, including Perplexity-powered answers for user queries.
The product’s reception made sustained use more important than launch interest: an early hands-on found basic AI queries workable but described the Rabbithole app as unfinished, while a prominent review criticized the broader pattern of shipping incomplete AI products with promised fixes. Lyu’s daily-user figure is therefore a concrete, if company-reported, measure against that backdrop.
First-order effects
- Rabbit now has a reported base of roughly 5,000 daily R1 users, giving the company an operating audience whose usage can inform fixes to the product experience that drew criticism.
- The figure shifts the immediate discussion from R1’s design and launch appeal toward whether Rabbit can retain users despite the criticism that the device was sold before it was fully ready.
Second-order effects
- Other AI-device makers face a clearer expectation to show recurring use, not merely preorders or initial sales, when software capabilities remain central to the hardware proposition.
- For Rabbit’s AI partners, daily engagement becomes more consequential: the R1’s value depends on reliable external services as well as the device itself, reinforcing the importance of complementary assets.
Third-order effects
- If early AI hardware continues to reach consumers before its software workflows mature, the category may be judged increasingly on post-launch iteration and retention rather than industrial design or launch-day demand.
- The R1 illustrates an unresolved category test: dedicated AI devices need to establish a durable use case that justifies separate hardware, rather than relying on the promise of future software improvements.
The trend: AI hardware is moving from launch-hype metrics toward proof of repeat use and dependable software integration.