The Rabbit R1, a Textbook AI-based Pump and Dump Scheme

The Rabbit R1, a Textbook AI-based Pump and Dump Scheme

The founder of the Rabbit R1 has done everything he can to hide the fact that it's basically just a tiny cloud-based API gateway between a $200 over priced badly implemented gadget and OpenAI.

In case you're not technically savvy enough to understand what that means, it's basically a device with a camera, a really bad screen, a really bad loud speaker, a small microphone, and a phone chip with a SIM card. There's no AI in the product, and at best it can be described as a badly executed custom GPT.

Whenever you speak to the Rabbit, it will record what you say, transmit it to Rabbit's servers, for then to transmit your input coupled with a system instruction to OpenAI. Then it will return whatever OpenAI returns back to it to you and speak out loudly. If you've ever created a custom GPT, you've basically created similar systems yourself, ignoring the Large Action Model parts of it (LAM).

Rabbit's LAM

The Large Action Model Rabbit brags about is actually just PlayWright scripts invoked on the device towards their API. The type of stuff we developers have been using to test our UI for decades. This implies there's zero AI in the process, and we've had similar features in Magic Cloud for more than a year using Hyperlambda - Except we build our actions on APIs and not by reverse engineering or recording browser clicks - Implying ours is 100x more resilient to changes than Rabbit's scripts. To understand just how bad the Rabbit R1 is, watch the following video made by CoffeeZilla about it.

Please, please, please watch the above video before you waste $200 on the Rabbit ...!!

Rabbit's NFT origins

Back in 2021 a friend of mine tried to sell me on NFTs. I tried to politely decline the offer by saying something roughly equivalent to the following.

I know absolutely all cryptography theory required to understand NFTs, in fact I know so much about cryptography that I could probably have been the author of the BitCoin white paper. However, you could explain NFTs to me for a million years, and I'd still not get it - So I'll pass on your offer

In fact the only crypto I have ever owned in my whole life I was given - And yes I made 5 digits on it, but I wouldn't buy a BitCoin if you told me I could get it for 1 cent. Because as far as I'm concerned that would be 1 cent wasted on a SHA256 hash value, and my computer can create 1 million of these per second.

However, and this is where you should pay attention. Rabbit as a company made its first money selling NFTs. If you believe NFTs have value, then by all means, go buy the Rabbit R1 - At which point you should probably stock up on 200 Rabbits, since they might increase in value in the future, right?

I can create my own PlayWright scripts though ... 😉

An (almost) honest attempt

If you're to compare the Humane to the Rabbit, then the Rabbit is inevitably better apparently, according to all reviews I have ever seen. This is an unfair comparison if you ask me, because the Humane pin at least did an honest attempt at creating a real product - While the Rabbit is literally the type of stuff you get if you put 50 teenage junior devs without any experience into a room, give them an OpenAI API key, and tell them to create an AI assistant.

Rabbit's code have been reversed engineered, and it's literally so bad, that none of the hackers even dared to reveal it because of all the security holes in it. I could make a qualified guess at what we're looking at though, and my guess would be hardcoded passwords, or complete lack of passwords, no SSL when contacting the cloud, transmitting passwords in clear text, hard coded passwords on the memory chip, etc, etc, etc.

To put their security issues into context, CoffeeZilla didn't even have the courage to turn on the device in his studio, because he didn't want the whole world to know where his studio is! 😳

I believe devices such as these are really great conceptually, but the problem is that it's at best "an app". Creating a separate device for a personal assistant is just madness, when it can be much better executed as simply an app for your phone. Combine this with the joke of Rabbit's "Large Action Model" (which is just 4 PlayWright scripts, dynamically executed by having ChatGPT constructing JSON payloads it sends to Rabbit's cloud servers) - While selling this is "a better LLM with action capabilities" is just plain wrong on so many levels.

It's a classic pump and dump scheme

However, and this is where you really need to pay attention. Rabbit comes without a subscription fee. This simple fact combined with that they're using OpenAI's APIs behind the scenes, implies that it's simply not financially sustainable. If you use your Rabbit device for 12 hours per day, for 3 months, then Rabbit will have to subsidise your OpenAI API tokens, and they will basically have spent more money on OpenAI API tokens than you spent to pay for the device.

This implies that inevitably after some few months of usage, Rabbit will simply have to turn off your device, unless they're willing to give you money for free

After 3 months of usage, their cost to operate your device would rapidly approach many times over the money you spent for purchasing the device. Implying they'd either go bankrupt or have to shut off your device, or sponsor you with "free money".

The last point is crucial because it's the defining factor of an AI-based pump and dump scheme. I've seen hundreds of similar schemes over the last year, and they all promise you "pay one up front fee, and use forever". The up front fee pays for your OpenAI tokens for the next 6 to 12 months, implying if they shut off your OpenAI access after 3 to 6 months, they've made 50% ROI on the purchasing price you paid, while leaving you without a product after 3 to 6 months.

At least the Humane pin is an honest product, and an honest attempt at creating a legitimate product, with a subscription fee and therefore a sustainable business model - Or at least an attempt at creating one. The Rabbit R1 is a scam. It has to be. It's simply not financially sustainable for Rabbit to continue paying for your OpenAI tokens in the long run ...

But then again the CEO of Rabbit claimed in 2021 that he'd be creating BitCoin 2.0, which would produce electricity during mining - So what do I know - I'm just a dev head with 42 years of experience in these matters ... 😂

Conclusion

If you want to have a legitimate AI assistant, we're creating a real "Large Action Model" these days, by utilising Hyperlambda and the plugin architecture of Magic Cloud. This will be based upon OpenAI's ability to generate code, structured data, and JSON - But we're not hiding that fact.

The Rabbit R1 however, is a rubbish product, created by rubbish developers, hired by a rubbish CEO, who is spending 90% of his day lying through his teeth, to attract idiot investors and retarded consumers - Sorry, you can sugar coat it any way you wish, but it's just the way it is ...

Thomas Hansen

Thomas Hansen I am the CEO and Founder of AINIRO.IO, Ltd. I am a software developer with more than 25 years of experience. I write about Machine Learning, AI, and how to help organizations adopt said technologies. You can follow me on LinkedIn if you want to read more of what I write.

Published 25. May 2024