The hype around Rabbit’s r1 AI assistant gadget may have cooled since its impressive debut at CES, but the company has rolled out some updates that might reignite interest—though they may not be enough to sway its skeptics just yet.

A newly introduced "beta rabbit" mode aims to boost the device’s conversational AI abilities, particularly when handling more complex or multi-step commands. This update is designed to help the r1 ask follow-up questions when it’s uncertain about a user’s request, offering a more interactive experience.

For instance, users can now ask:

“Beta rabbit, can you suggest three books similar to ‘The Power of Now’? Include page length, year of release, and ratings, and save that as a note titled ‘reading list’. Also, include pictures of the authors.”

This could be followed by:

“Beta rabbit, can you also get me summaries for those three books?”

The new features also extend to setting travel itineraries and finding deals or product recommendations, although these functionalities may still fall short in real-world usage. As experienced chatbot users can attest, these tasks often work well in demonstrations but prove less practical in everyday scenarios. Travel itineraries can still be somewhat erratic, and product comparisons—though technically impressive—might be inconvenient to perform on such a compact device. Furthermore, users may hesitate to rely on AI-sourced book recommendations that could lack credibility.

The update also includes enhancements to alarms and timers, which are certainly welcome but occasionally lead to impractical results. For example, asking the r1 to “set a timer for baking chocolate chip cookies” may raise questions: at what temperature? How many cookies? What type of chocolate chip cookies? These are essential details the AI might not account for. However, it would be reasonable to ask it something like, “How long should two dozen chocolate chip cookies bake at 300 degrees?”

Despite these updates, the most anticipated feature—the much-discussed yet still elusive "large action model" (LAM)—remains out of reach. Promised back in January, the LAM was described as a system trained to autonomously navigate phone and web app interfaces to execute tasks selected by the user. To date, this capability has not been demonstrated beyond controlled demos. Any actions claimed to utilize the LAM are indistinguishable from what could be achieved with standard APIs or scripted actions.

While the potential of this quirky little gadget remains intriguing, its full utility has yet to be realized. This is why, despite its limited use since receiving it for review, I haven’t completely given up on the r1 just yet.

I’ve reached out to Rabbit for an update on the LAM’s release and will provide further details if they respond.