Is the Rabbit R1 a Scam?! (Not really)

I haven’t personally used the Rabbit R1 device so far. All my information comes from second-hand reviews and Rabbit’s own press publications. It’s undeniable that the R1 has met some criticism (rightfully so) for overpromising and underdelivering on features.

If you’re not familiar, Coffeezilla is a great investigative YouTuber with very entertaining and downright brilliant videos on scams and malpractices in various tech sectors. However, there’s a growing trend of claiming scams on things that could simply be attributed to a bad go-to-market strategy.

Coffeezilla’s Critique

I saw two videos by Coffeezilla on the R1 device. The first breaks down the career of the company’s CEO, the company identity, and its vision over the past few years. In a follow-up video, the YouTuber makes the argument that the device runs on ChatGPT-3.5, not a proprietary LLM, and uses an automation layer through another third-party software, Playwright, to navigate supported apps.

Oversimplified diagram of "LAM"

The Large Action Model (LAM) is actually a crude hardcoded path tree to different predetermined functions. This means the R1 is constantly wrong when answering present-day questions and frequently misses the mark when trying to perform tasks. Coffeezilla argues that since the device was marketed as a virtual assistant that can adapt and respond to your everyday queries, the product is a scam.

Product Launch and AI Market Perspective

Here is where I strongly disagree with Coffeezilla. Are the R1’s complaints justified? Absolutely: the hardware might be cute and ooze Teenage Engineering goodness, but the software and functionality are unpolished, inconsistent, and downright useless.

But a scam? No.

Rabbit raised capital for their AI product at the end of 2023. They needed to come up with a value proposition for users and validate it with a product. This product had to be easy to develop in a short period of time. After all, they knew they’d be competing against giants like Google and Apple, who could integrate a better version of their product into their phones, which are already massively adopted. They had to do all this before Q3 2024.

Rabbit also needed to stand out and captivate an audience before it was too late. From a product perspective, I must say they addressed a lot of these points in a pretty genius way:

• They partnered with Teenage Engineering to create a unique gadget and borrow their hardware expertise.
• They made it way cheaper than the closest competitor: $199 for the R1 vs $699 + subscription for the Humane Pin.
• Instead of building their software from scratch, they partnered with existing products and built their own “wrapper” product.

Wrapper products take features from one or more existing products and create synergies that serve more specific use cases. This is common, not only for MVPs but even for big players. Apple is negotiating with Google and OpenAI to see which company will power their new AI program.

Where Rabbit Went Wrong

The R1 does a decent job creating a good wrapper MVP for their “Large Action Model,” combining GPT-3.5 and some AI ops automation with Playwright to mimic their own “intelligence.” However, Rabbit failed in two critical areas:

1. They neglected to narrow down their target audience, making their product too broad and promising what they thought the general tech enthusiast wanted to hear.
2. They did a poor job of communicating the limitations of their system, trying to present more than what they could actually accomplish. Compare their announcement with OpenAI’s announcement of GPT-4. OpenAI allows mistakes to be part of the announcement, acknowledging hiccups and even failures, setting realistic expectations for the audience.

What Would I Do?

Hindsight is 20/20, and the internal and external pressure the team was under was probably a big part of why they rushed their discovery and launch process. Here’s what I would have done:

1. Identify an audience interested in optimizing everyday tasks with AI. Focus on developers, makers, product managers, social media managers, etc. These early adopters expect unpolished items and can’t wait to improve them.
2. Underpromise and overdeliver. Change the narrative and ship with fewer features, but let the community know they can improve the automations: “We’ll create the LAM together.”
3. Make the LAM open. Embrace transparency and engage developers by creating an SDK to develop their own automation (providing access to Playwright, Zendesk, etc.).
4. Launch Rabbit extensions. Create a webstore with extensions for Rabbit, allowing developers to post and rate them. Rabbit could start their own app store.

This approach might have taken more time and planning, and initial sales might have been less impressive, but it would have saved Rabbit from the PR nightmare they’re in now. They are not a scam, but a case study in lack of foresight and poor launch strategy.