Rabbit R1 AI Agent

Definitive article about the design and the “couldn’t it just be an app” comments.


Good article. I do think it has been very well strategically conceived from hardware to business model.

1 Like

Obviously I should have signed me up. Did not see a “sold out sign” comming. 50k Shippments for this seem to be on the very conservative side.

1 Like

I just got an ad on IG for the r1. I almost bought it. Seems there’s a new release of more units shipping in June I think.

Still don’t know what I would actually do with and can think of a few other things I’d rather have for $300…

I feel like everyone forgot about this thing that everyone was talking about last year.

Ok, so I’ve been thinking about this thing a lot. I think it’s a cool concept but I can’t imagine it working, even putting aside the hardware.

If the AI is based on tasks and can learn a UI once you connect a user login or API, don’t you need to sign into every service and ensure that plugin works? How does it handle services that require a new login or user account?

Most tasks that I imagine you’d want this thing to do would be the ones that you don’t do very often but are time consuming. Booking a vacation, etc. If you need to teach it the UI for that app/website doesn’t that defeat the point?

Even in the demo, it seems that everything needs confirmation by the user. That’s not helpful. If I had a real assistant and told them to book a meeting and get me lunch, I would be super annoyed if they came back for an “OK” for everything - is this the room? Is this the meal? Do you want a drink? I can’t see how this could know enough to do it all.

The demo also showed a lot of responses with useless background info. The cast of a movie or something. If I want to book a movie ticket, I don’t care about where the director grew up, just do it.

IF this would work it would mean that nobody looks at a website. That means no traffic to see/click ads. I can’t see the big players letting that happen. How easy would it be to ensure that this use case is blocked like how Youtube is making it run slow for browsers with ad blockers?

The biggest challenge with this or any AI is conflicting interests. The user vs. the platform. The user wants to get things done efficiently and see only what they want. The platforms, most of which are ad or service revenue supported don’t want that. They want you to see what they want you to see, be interrupted with ads, and buy what they are selling. Any AI trying to be smart enough to edit out those ads or only show a user what they want will be confronted with a stronger AI made by a company with more resources to fight that. The AI assistant that is unencumbered by ads and finds the best price/product/selection based on what you actually want will never happen.

Good reminder. Some clear industrial design lineage on the R1 with the Playdate. A parallel to the enthusiasm between the two seems less apparent. The PlayDate was a quirky, purposely retro one-off with no paradigm-changing ambitions. I read the release media, looked at the crank, wondered, but never thought to order, spoke about it, or wrote about it. No one ever mentioned it in my circle, but pocket gaming consoles are not me. Google Trends shows about 2x the impact of the launch announcement for the R1.

The reviews are uniformly good. “Charm” seems to be the operative word. They sold 53,000 pieces of the PlayDate in one year, as of April 2023, not bad. Open system for developers, 400 games released for it, good fun object. Wish them all the best.

Okay now I understand the crank. Seems cool. 400 games released. Open.

I have 200$ worth of skin in the game having ordered an R1 on launch day. I have been thinking of what I will do when it arrives. That process of questioning the status quo of net interactions is already worth the price of the ticket. I resist the ossification of ideas and approaches, interaction also involves embracing change and innovation to avoid the rigidification of thought.

This R1 device, that keynote, brought front and center the odd little silo of interaction I have accepted. It seems intriguing to try an AI device that will try to get under the advertising layer of the Apple and Android walled gardens.

Although the R1 is technically beyond my expectations of what is possible now, I hope to be surprised when the box gets opened and when people start working with the tools.

Potential case example:
My sister is taking my late mother’s lifetime of collectibles and trinkets and looking to productively disperse them on various online platforms, as well as make a heirloom picture catalog for the family.

The process involves:

  • Taking pictures
  • Writing engaging descriptions about the object in the image (does this like magic)
  • Adding search tags and SEO
  • Uploading to various marketplaces
  • Writing tags into the image metadata
  • Tracking the process
  • Repeating these steps for each unique item

Currently, I can open a chat in Perplexity or chatGPT on the phone, take or upload a photo, optionally record a short description, and ask it to generate a full descriptive Etsy or eBay listing. It does an amazing job at this. However, it cannot then automatically generate such a listing. It cannot leave the confines of its application.

Someday soon this will be integrated into a single marketplace’s structure, but it does not exist yet. Yes, there are automation APIs, and building blocks with Node-RED that I could assemble a solution, but the idea of the R1 opens this door, and I hope it can do this chain of events.

The other aspect is that this device will be a conversation starter, and with conversation comes new ideas. Just as the white finish on the Apple portable products acts as a highlight to attract attention, the bright colors of the yellow PlayDate or the orange RI do the same. Not a new idea, Sony Sports Walkmans employed the same strategy of visibility highlighting. I am looking forward to the discussions and the new approaches that emerge, even if the R1’s function gets crushed or hindered by the ad revenue and ecosystem giants.

1 Like

Well put.

This is like the first Mac -novel form factor, approachable feel, new UI, consumer and use friendly use of tech and not so much use at first, but lots of potential.

These astroturf products are very difficult to make stick.

I worked on a similar project a few years ago that made similar AI functionality promises. What we learned from investors is that any AI startup brand entering this space needs to show they know how to do hardware in order to be taken seriously.

The R1 is now poised to explore their early stage startup claims on the backs ($) of exuberant early adopters willing to lose $199 in order to raise the next round of capital that will test their claims made in this current campaign. Hence the simple primitive form factor concept representing fulfillment of hardware capability by the startup team. They are basically ticking boxes for the investors with this hardware concept form factor.

The big claim here is the ability to tap into other companies databases in order to perform all matter of AI tasks. I don’t see this happening with larger legacy data companies. (too many data security protocols to surmount) Rabbit Tech needs to target other similar new startups and partner with them as they and the broader AI task agent sector grows.

my understanding @designbreathing is that this doesn’t really connect with other platforms or apps through APIs or infrastructures but rather acting as a user via running a version of the same software or login the user wants but remotely? Maybe I misunderstood. ie. rather than work as a plug in to Photoshop, if you ask it to do PS it will just do the task running PS on a server elsewhere and then send you the results, or if you want it book an AirBnb it will user your login on the actual website like a human would navigating the UI to do it vs. working like a machine through some computer protocol.

I have had a good think on how “astroturf” applies to this product. Artificial grassroots support? Faked product to build investor enthusiasm? Still drawing a blank. Other products as examples?

The shipping date is close, we shall see. The $199 “gamble” has already resulted in the gift of an equal value Perplexity subscription, which I am using every day, its good.

“Hence” is doing a lot of heavy lifting here. I think the design is amazing, the choice of bringing in TE was great, not a shortcut to check a box for the suits. The near-design sibling PlayDate is described by all reviewers as charming. Charming is a good way to mask the threat of the AI future, :face_with_spiral_eyes:

At the speed of growth, what was not possible six months ago is possible today. I am not an AI evangelist, but I see the writing on the wall. I realize that for decades people have worked on AI projects. A lot of things are happening now that were impossible a few years ago.

I did not get the impression from watching the keynote that other databases would be tapped into in a way to trigger security, just clicking on the frontends to automate tasks. Might watch it again.

At 11:45 of the demo he’s talking about the Rabbit Hole feature. “And of course we do not store any of your third party credentials.”

If you fail to see the bait and switch here, you’re not understanding how silicon valley works to try and create AI innovations on the backs of legacy databases. Their feature must remember login credentials in order to access third party apps, therefore they must 'store" it.

It’s like looking directly at the face of the People’s Liberation Army when he says that. Ha!

This issue is currently being discussed in the courts of Europe under GDPR statues. Italy in particular. ChatGPT is violating Europe's privacy laws, Italian DPA tells OpenAI | TechCrunch

This product targets GenAlpha and parts of GenZ hoping to capture their data and their loyalty early in life. Time will tell if this generation of users is as easily swayed as the millennials were to give up their privacy.

Can’t it somehow use locally stored login info to access a remote login via some sort of key? Like how online shops don’t store credit card info but can process it?

I think claiming this whole brand, AI software model and brand new hardware device is just a play for user info is pretty far fetched.

1 Like

From the creator, here is a great interview. Smart man, has a real vision. He has been on the board of Teenage Engineering for years and fluent with both design and experienced in approaches to the app “problem”. Lots of insights and answers into the speculations raised here. Chapters marked.

01:03:48.039 Another part is that phones are just too boring now, like industrial design is pretty much dead

I share his opinion on the boring state of design we have arrived at. Everything feels stale and predictable. I get that slabs do what they need to do. Even grandfathers can write tweets all day long without usability issues.

Wow. :grimacing:

I worked hard, lived well, partied, started businesses, thrived (except for the AQI), and started a family in the Communist China city of Shanghai over 20 years. Glad to be out, but the experience I would not change.

Surveillance or the threat of it, is weird when conducted by any nation, trillion-dollar company, or eccentric, erratic, South African billionaire. The Rabbit R1 is just a device that plays on the same playing field as every other device. Demonization of the R1 on this front seems, odd.

This is where we operate now, no one forces us to. We can choose to quietly consume information, use tools to become more adept and share, or overshare. The R1 approach is a cross-generational product with as many nostalgic design references as minimal modern naive-projecting directions.

Surprisingly to me, the R1 emerges as some sort of litmus/Rorschach test, which seems worthy of study on its own.

The “problem” is that no one called themselves a designer, not even the elusive designer.

But hey - I’ve called myself an engineer when it suited me as well. Only because I have dual degrees though, I wouldn’t dream it otherwise. And I sure as hell never called myself a psychologist, anthropologist, ergonomist, business analytic, marketer or anything else I might do on a daily basis for past 20 years, but not have a degree in.

Are you saying anyone who designed a thing that struck a chord with an audience can call themselves a designer?

I’m an industrial designer, graphic designer, footwear developer, marketer, business person…

I guess only losers pretend to be designers if they don’t have an ID degree?

Dieter Rams began his studies in architecture and interior decoration at Wiesbaden School of Art in 1947, now part of the RheinMain University of Applied Sciences. A year later, in 1948, he took a break from studying to gain practical experience and finish his carpentry apprenticeship. He returned to the Wiesbaden School of Art in 1948 and graduated in architecture with honours in 1953

1 Like

WaPo had a piece decrying the Apple goggles as intended to do exactly that, with accurate maps of everywhere you use the product.


Apple: Let’s just record a 3D scan of your face and then algorithmically reanimate you in Personas!

Apple NDA’d internal marketing: We think we should tweak the smile settings a bit to reflect more accurately the communicated joy of the user experience. “We think you should smile more.”

I really don’t understand the obsession with weird surreal avatars. Memoji, the thing Zuck was trying to do with Metaverse.

Do adults actually want this? Hasn’t Zoom fatigue taught us that we don’t always need to see our self/others?

Just because you can…

1 Like