[HN Gopher] Rabbit R1 source code [part 1]
       ___________________________________________________________________
        
       Rabbit R1 source code [part 1]
        
       Author : wibbily
       Score  : 292 points
       Date   : 2024-04-23 18:18 UTC (4 hours ago)
        
 (HTM) web link (github.com)
 (TXT) w3m dump (github.com)
        
       | ado__dev wrote:
       | Not surprised at all. This and the Humane Pin both seemed like a
       | quick cash grab before phones integrated all the new AI goodness.
       | I'm expecting we'll see that window close by the time I/O and
       | WWDC wrap up this year, so they had to move fast.
        
         | pnw wrote:
         | A company founded in 2018 that raised over $200m is a "quick
         | cash grab"?
        
           | SheinhardtWigCo wrote:
           | "Quick" as in being quick to exploit LLM hype.
           | 
           | "Cash grab" as in Elizabeth Holmes grabbing $700m from
           | Theranos investors.
        
         | SheinhardtWigCo wrote:
         | Their "keynote" just _screams_ shenanigans. It 's the most
         | blatantly fake product demo I have ever seen.
         | 
         | To take one example, at 14m30s, the CEO is shown using the
         | device to book a trip to London. This is presented as a live
         | demo, but it's clearly simulated.
         | 
         | https://www.youtube.com/watch?v=22wlLy7hKP4
        
           | imiric wrote:
           | Playing devil's advocate, tech demos are often faked. From
           | the famous "Hello" Mac introduction in 1984[1], to Google
           | with their Gemini demo a few months ago.
           | 
           | A certain degree of smoke and mirrors to generate hype around
           | ground-breaking tech is the norm, not the exception. This
           | doesn't necessarily mean that the product itself is a scam.
           | 
           | [1]: https://www.folklore.org/Intro_Demo.html
        
           | RockRobotRock wrote:
           | To be fair, he doesn't "book" a trip to London in the
           | presentation, he gets flight, hotel, and car rental
           | information, which isn't that out there.
           | 
           | But, he does imply that you _can_ book from the device, which
           | is completely insane. How do you deal with flight selection
           | with multiple layovers? Seat choices? DOBs? KTNs? Phone
           | numbers? Frequent flyer number? Payment? Travel insurance?
           | Disability accommodations?
           | 
           | What could possibly be the overlap of people who travel
           | enough that this is all worth setting up, who don't already
           | have their own personal assistant?
        
             | kmlx wrote:
             | just the payment part is it's own universe of issues,
             | friction, fraud, security etc etc etc
        
         | hncel wrote:
         | I don't think this is a fair characterization of Humane. I
         | don't and haven't worked at Humane, but I did interview there
         | and have some friends who work there now. They are notoriously
         | secretive about their product (founders are ex-Apple, and they
         | try to keep an Apple like secrecy culture) but I do know a bit
         | about the evolution of the product.
         | 
         | Humane was founded in 2018, well before ChatGPT was released in
         | November 2022. If you look online you can find some articles
         | about patent applications they made well before ChatGPT was
         | released that give you an idea about their idea for the product
         | at the time, e.g. https://9to5google.com/2022/01/07/humane-
         | android-ar-wearable...
         | 
         | Developing the hand tracking, laser projection system, voice
         | recognition, etc. is very hard, especially considering the
         | power constraints on the device. They spent years working on
         | this and when LLMs hit the scene they realized that the
         | original product idea was going to be severely lacking if they
         | didn't integrate this technology. This caused a big internal
         | pivot to more closely integrate with these LLMs. I'm not sure
         | which they're using, presumably they're paying for GPT-4 access
         | or something like that. It's understandable why they felt like
         | they had to do this, and why it feels like a rushed
         | integration. The bottom line is that they were way too
         | optimistic with the hardware capabilities when they started
         | working on the product, and the last minute rush to integrate
         | with LLMs to at least improve the software capabilities to kind
         | of close the gap is what we're left with. It's not a great
         | situation, but I also think it's unfair to characterize it as a
         | "cash grab".
        
           | megaserg wrote:
           | Secretiveness in a startup at that stage is not a good sign.
           | See e.g. Theranos or Magic Leap.
        
           | rodiger wrote:
           | Cash grab from VCs is probably more accurate... I have zero
           | doubt some incredible engineering has gone into their
           | product.
           | 
           | However, from what I can tell they were searching for a
           | problem to solve instead of coming with a distinct,
           | compelling, articulable vision of what they wanted to build
        
           | fragmede wrote:
           | the hardware is phenomenal, but it's tied down by bad
           | software
        
       | georgehill wrote:
       | Why did they upload the code to some random site instead of
       | GitHub?
       | 
       | EDIT:
       | 
       | > But let's call a spade a spade - this is a blatant lie. And
       | we're about to expose it with the first partial release of the
       | source code for its so-called "large action model".
       | 
       | FYI, Text to Action is possible. I personally tested a couple of
       | apps, but I don't think anything reliable exists like we humans.
       | 
       | I would not disregard what they claim is completely false.
        
         | paxys wrote:
         | Because it will be taken down from Github.
        
         | claytonjy wrote:
         | to reduce the chances of GitHub taking it down, perhaps? not
         | sure it'll help though
        
         | no-dr-onboard wrote:
         | It buys time for them. GitHub will have a lengthy internal
         | discussion about DMCA takedown requests and the fact that the
         | code isn't on their server. During this time it'll give the HN
         | and reddit communities time to grab the link and redistribute.
        
           | georgehill wrote:
           | Understood!
           | 
           | But zipped files are super fishy though. What if this repo is
           | spreading malware?
        
             | fwip wrote:
             | Zip is just a tarball that Windows understands.
        
         | dxbednarczyk wrote:
         | Potentially due to the threat of DMCAs. Pixeldrain and Mega are
         | widely used in piracy and sometimes leaks like these,
         | considering they are not known for complying with them most of
         | the time, unlike GitHub.
        
       | rodiger wrote:
       | At least it'll look nice on a shelf thanks to Teenage
       | Engineering's good work :)
        
         | WanderPanda wrote:
         | TE might have their name really damaged by this. I wonder why
         | they agreed to this colab in the first place
        
           | micromacrofoot wrote:
           | they were paid money in exchange for services
        
           | BadHumans wrote:
           | I don't think any fan of Teenage Engineering is no longer a
           | fan because they were paid to design a product that flopped.
        
             | serf wrote:
             | why not? they clearly wanted to exploit AI hype in order to
             | turn a profit, even if they did so indirectly. Why
             | shouldn't that speak to their motives and trustworthiness?
             | 
             | if Apple started churning out guns, landmines, snakeoil,
             | cancer cures, NFTs, and magic-AIs their reputation would
             | falter.
        
               | wilsonnb3 wrote:
               | Apple might be able to get away with all that, their
               | reality distortion field is unmatched, but I agree that
               | teenage engineering might actually take a reputation hit
               | if they are too liberal with their outsourcing.
               | 
               | Personally, the Vision Pro wasn't really my cup of tea
               | but I will be standing in line on day one at my local
               | Apple Store for the Landmine Pro.
        
               | kevindamm wrote:
               | Wait til you find out that the Vision Pro can detect the
               | presence of Landmine Pro in your vicinity.
               | 
               | It's all about that vertical integration.
        
           | chambored wrote:
           | I highly doubt that. TE has a strong reputation and their
           | involvement in another company's product outside of their
           | normal business won't impact their sales regardless of
           | whether or not R1 is a flop or a ruse.
        
           | talldayo wrote:
           | Selling underpowered toys at eye-popping prices has been
           | Teenage Engineering's modus-operandi since the beginning.
           | After the OP-1 price hike (and subsequent re-release at an
           | _even higher price_ ), I don't think their target audience
           | cares past this point. You either buy TE stuff for the brand
           | recognition or you own a bunch of Behringer gear because
           | you're poor.
        
             | vegadw wrote:
             | Honestly, while I agree there's a massive price
             | discrepancy, I don't know that I even see Behringer as the
             | "lower end" option anymore. The lowest end is a laptop and
             | pirated VSTs. Behringer pedals? Yeah, those are because
             | you're poor (or just don't care, I guess?) but synths? The
             | K2 is better built than the MS20 mini, the Wasp reissue
             | doesn't have a good equivilent I know of, and their 303
             | clone isn't any better or worse than the other similarly
             | priced 303 options - and it's easily hackable.
             | 
             | I mean, don't get me wrong, I still think Behringer as a
             | company is doing bad things to the music land scape and
             | that they've done some pretty horrendous IP theft and
             | racist stuff, but I don't think "because you're poor" is
             | right either.
             | 
             | To the point though, yeah, no, TE absolutely won't get a
             | black eye for this: Nobody cares, and hardly anyone but
             | tech bros even know. It's no worse than their wooden choir
             | thing.
        
             | caseyy wrote:
             | https://teenage.engineering/store/field-desk/
        
               | talldayo wrote:
               | I retract my previous statement. Indeed, toys are not the
               | only expensive stuff they have for sale.
               | 
               | Hey look down there! The computer-1 case is on sale for
               | $149, down from $249. How much profit you still think
               | they're making, considering it's a DIY kit of bendable
               | sheet metal?
        
             | noman-land wrote:
             | While I agree this has been true of TE for most of their
             | existence, their most recent $300 sampler bucks this trend
             | considerably.
             | 
             | https://teenage.engineering/store/ep-133/
             | 
             | Neither that, nor this little AI cloud device are eye
             | watering in price.
        
               | neom wrote:
               | People loooove to complain about TE pricing. OP-Z is
               | $499, find me anything else that can sequence sound,
               | video and DMX (on the go) like OP-Z can, there are
               | exactly zero others. OP-1 for portable synths.... are
               | people pulling a Deluge out on a flight? I've seen people
               | compose a whole set with a OP-1, over 30 minutes, with
               | nothing else... good luck doing that on a MC-101.
        
               | talldayo wrote:
               | I own a Pocket Operator, my head's not in the sand about
               | their cheaper offerings. The Pocket Operators are cheap
               | toys though, you cannot tell me with a straight face that
               | it's about to prop up your next EP. It's a mass-produced
               | calculator PCB they forgot to make a case for, and sell
               | at insane markup.
               | 
               | The same goes for the KO and indeed, the OP-Z. The OP-Z
               | doesn't even have a screen, it has no business costing
               | $499 for being a bunch of buttons with a USB-C plug.
               | 
               | > find me anything else that can sequence sound, video
               | and DMX (on the go) like OP-Z can
               | 
               | How about the mandatory paired device it requires to
               | sequence everything? That iPhone/iPad is certainly
               | capable of doing that itself, alongside multiple things
               | the OP-Z _can 't_. Nevermind how far you'd get with a
               | $300 laptop and $200 DAW.
        
           | j4ustin wrote:
           | The Rabbit CEO is on the board of TE.
        
         | slily wrote:
         | This reminds me of how Playdate owners make it very obvious
         | they don't use the thing in the way they retroactively justify
         | their purchase by commenting on quirkiness or aesthetics at the
         | expense of functionality or usability. I guess I get it but
         | there's cheaper plastic toys out there.
        
           | mulderc wrote:
           | I don't think that is retroactive, people bought it because
           | it was different and looks cool and is quirky. I think it is
           | a very fun device and has some great games for it. It isn't
           | for everyone but it is a cool device for people that
           | appreciate what Panic is doing with it.
        
         | kotaKat wrote:
         | I bought it because I figured out it's gonna be some crummy
         | little Mediatek thing underneath the skin (and it is) running
         | some form of Android (and it is) so I'm just sitting back to
         | hack it back to be a real communicator ;)
         | 
         | EDIT: It's an MT6765 (Helios P35). It's got a known BootROM
         | exploit. Won't be long until someone dumps it and cracks it
         | open, though would be hilarious if a part2/part3 dump is just a
         | factory stock ROM.
        
       | aw4y wrote:
       | I expect it (or at least I hope) to be really hackable.
        
         | plugin-baby wrote:
         | Based on how hackable their infra is?
        
       | rvz wrote:
       | Big if true.
       | 
       | Could set back the AI device hype 5 years back after Humane
       | getting exposed as another scam.
        
         | jagger27 wrote:
         | You already have the best AI hardware we'll see for a while in
         | your hand.
        
           | passion__desire wrote:
           | I remember Mozilla was working on 3D web browsers. Could be
           | new way to do things in Vision Pro? Is there any progress on
           | that front?
        
       | adlpz wrote:
       | Any background on how the code got leaked? Insider? Hacked
       | servers?
        
       | jthnme wrote:
       | Some will call it scam, some will call it MVP
        
         | godelski wrote:
         | If your product doesn't work anywhere near what your "live
         | demo" shows, it's a scam.
         | 
         | That's very different from "here's the product we envision and
         | need money to build it."
         | 
         | And just because others have scam demos (including Gemini) that
         | doesn't make it okay. It makes it a race to the bottom (and is
         | why I'm more upset about Gemini because big players are held to
         | higher standards)
        
         | ikurei wrote:
         | In and of itself, the product might be a decent MVP to validate
         | the idea or some aspects of the design.
         | 
         | The problem is in how they've marketed. If you're taking
         | people's money and giving them an MVP, you need to be upfront
         | about it; if you aren't you're doing a bad thing.
        
       | paxys wrote:
       | Meh. I don't think they ever hid the fact that the device is
       | basically a ChatGPT wrapper. As long as it can achieve what it
       | advertises, who cares how the backend looks? At least it has the
       | decency to charge a reasonably price ($200, rather than $700 +
       | subscription like the Ai Pin).
        
         | furyofantares wrote:
         | > I don't think they ever hid the fact that the device is
         | basically a ChatGPT wrapper.
         | 
         | "Large Action Model"
        
         | k8svet wrote:
         | >As long as it can achieve what it advertises, who cares how
         | the backend looks?
         | 
         | so if I pump out enough advertising, you're going to give me
         | the usernames, passwords, and active sessions for your accounts
         | to me?
         | 
         | I need to log out of this thread asap. I thought the defenses
         | of Ai Pin were going to drive me nuts, I need to preserve some
         | sanity. Has everyone lost their minds? Are tons of people here
         | working for equally scummy, shoddy, if not scammy, startups?
         | Seriously, what the hell.
        
           | paxys wrote:
           | > you're going to give me the usernames, passwords, and
           | active sessions for your accounts to me?
           | 
           | You give all of those to every smartphone maker. Why is this
           | any different? Is there evidence that their handling is
           | insecure?
        
             | k8svet wrote:
             | If I found out that Android was eavesdropping my Spotify
             | credentials, I'd be just as stupified, yes.
             | 
             | If I found that Android built in some Spotify integration
             | that worked by stealing my active session cookies to do
             | some backdoor integration with it, and billed it as some
             | future AI smart service, I'd find it equally g-d absurd,
             | yes.
             | 
             | Do I think that me logging into the Spotify app, in
             | Android, and it exchanging those credentials for an app-
             | internal access token is the same as a server hijacking my
             | session? No, not really, I don't.
             | 
             | That's what's so damn brazen and shoddy about this. SPOTIFY
             | HAS OAUTH.
        
               | paxys wrote:
               | I have no idea what you are trying to say. The device
               | works by running apps for Spotify, Uber etc. in a VM and
               | logging you into it. They say it right on their homepage.
               | If you don't trust it, sure don't buy it. That's your own
               | decision, but doesn't make them any more right or wrong.
        
               | k8svet wrote:
               | I'm saying it's shoddy, and scammy, and I can't believe
               | anyone would lift a finger to defend this type of
               | product, engineering, or actively training people to get
               | phished. Hope that clarifies.
        
             | ado__dev wrote:
             | Yeah, I have a lot more faith that Google and Apple will
             | properly secure my private data vs a random startup.
        
             | madeofpalk wrote:
             | Well, I actually don't. I only use hardware from companies
             | that I have a semblance of trust in, and I certainly don't
             | run around entering my Spotify or Uber password into other
             | services.
        
           | bogwog wrote:
           | This is probably what it felt like to be alive during the
           | peak of the "dotcom" craze I read about in historical
           | literature.
        
           | cogman10 wrote:
           | Haven't you done that already?
           | 
           | I mean this in all seriousness, have you used Oauth with
           | google/facebook or the like to login and register with online
           | services? Why not? Have you put passwords into a password
           | manager? Why?
           | 
           | Did you give Uber or Lift your credit card number? What if
           | they were a scam?
           | 
           | I say this also thinking rabbit R1 is a pointless product
           | that based on hype that nobody should buy. However, I can see
           | why people might think it reasonable to give their AI
           | assistant a bunch of personal information. For the same
           | reason people have trusted google with health data.
        
             | k8svet wrote:
             | > have you used Oauth with google/facebook or the like to
             | login and register with online services?
             | 
             | No, I don't use federated login _anywhere_. I can show you
             | my Google account. The only place I 've compromised is
             | Tailscale, and I plan to replace that imminently. And
             | frankly I consider it lazy of them to not support email,
             | especially since google.com accounts are single-tenant
             | anyway. _And tailscale never sees my password, never has
             | raw access to my entire damn account, etc, etc_.
             | 
             | Also, besides, federated login or delegated access, sure,
             | OAuth is great, I wouldn't have commented in this thread if
             | they were using it. Typing my raw creds into a [redacted]
             | VNC session is not comparable.
             | 
             | >Did you give Uber or Lift your credit card number? What if
             | they were a scam?
             | 
             | I call my credit card company. They reverse the charge, and
             | ding the merchant. My life goes on. Takes a shockingly
             | small amount of time.
        
         | skywhopper wrote:
         | It can't achieve what it advertises, though. I mean, what even
         | did it advertise? Voice transcription? Playlist management?
         | Phones do these already. "Order me a pizza, whatever the most
         | popular option is"? That's laughable. No one actually wants
         | that.
        
       | GaggiX wrote:
       | We already know that the large action model would not be
       | available at launch, but I wonder how well it does works as an AI
       | assistant.
        
       | godelski wrote:
       | Please upload source in an unzipped format. If the concern is
       | about GitHub taking it down, use an alternative. There's plenty
       | and many other ways to distribute source in an uncompressed
       | manner.
       | 
       | Otherwise this is indistinguishable from a hack. How do I know
       | these zips are secure? The mega and pixeldrain report different
       | sizes. Rabbit is entirely about hype and a scam, how are we
       | supposed to know this isn't the same nefarious ploy?
       | 
       | I appreciate what's being done and think it's good to call out
       | these scams (I've done so myself) but help by building some
       | trust. We understand the need for anonymity but a nefarious actor
       | could just as easily mascaraed as the same repo. And if you do
       | need files downloaded, provide hashes.
       | 
       | (Fwiw, xz, despite recent events, is great at compression and can
       | help you reduce your bandwidth if needed)
        
         | titaniumtown wrote:
         | > (Fwiw, xz, despite recent events, is great at compression and
         | can help you reduce your bandwidth if needed)
         | 
         | zstd level 22 is even better in my experience
        
         | nebulous1 wrote:
         | What is the reason for it to be uncompressed?
        
           | mimischi wrote:
           | I suppose the risk of a 0-day in the compression format,
           | given we're in the post-xz-era. Publishing the source code in
           | clear text would alleviate such risk for the consumer
        
           | godelski wrote:
           | 1. I can read it on my phone or in my browser.
           | 
           | 2. Why should I have to download text to __read text__?
           | 
           | 3. We don't want to normalize unnecessary behavior that is
           | something scammers and bad actors can easily take advantage
           | of.
           | 
           | While I don't believe the leak is nefarious or contains an
           | exploit, normalizing a requirement to download files that can
           | issue exploits -- when there are easy alternatives that make
           | this unnecessary -- just helps create the exact type of
           | environment that scammers thrive in. 3 is incredibly
           | important. If we're going to call out scammers we shouldn't
           | do it in a manner where we're enabling an environment for
           | more scammers to thrive in. Doing what's done here just
           | created a rich opportunity for hackers who can now post a
           | "rabbit source code leak" and just provide people with a
           | different link. Makes for easy picking. Uncompressed and
           | readable code just makes this harder and easier for people to
           | determine if something nefarious is going on.
        
             | nebulous1 wrote:
             | It's not a single file unless it's tar'd or compressed or
             | whatever. It's completely normal to distribute software
             | projects as some form of archive. This is doubly true for a
             | "leak" like this where you want the single file to spread
             | around.
             | 
             | I agree that it would be nice to have it browsable online,
             | like in a github repo or whatever, but that's a separate
             | issue.
        
               | godelski wrote:
               | > It's completely normal to distribute software projects
               | as some form of archive
               | 
               | Again, I think you're missing my point
               | 
               | >> normalizing a requirement to download files that can
               | issue exploits -- when there are easy alternatives that
               | make this unnecessary -- just helps create the exact type
               | of environment that scammers thrive in
               | 
               | Yes, it is "normal" and that is exactly the problem.
               | 
               | Ask yourself this                 Is there a reasonable
               | alternative?        Is downloading necessary?
               | 
               | I think you'll find that the answer to both is
               | unambiguously "no." I think you'll also recognize that
               | having the readable source __also__ unambiguously creates
               | higher utility.
               | 
               | So you don't need to explain to me that this stuff is
               | normal because I already understand that (and am actively
               | demonstrating a knowledge of this). I realize
               | communication isn't always obvious, but if someone is
               | telling you that you're missing the point of what they're
               | saying, please consider that you might actually be
               | missing the point rather than doubling down. Even if you
               | aren't, someone telling you that indicates that somewhere
               | there's a miscommunication, and that needs to be
               | resolved.
        
               | nebulous1 wrote:
               | I would prefer that it is distributed as a zip. It allows
               | me to easily get the entire file, and hash it to make
               | sure it's the same file as other people are getting, and
               | have an archive of it.
               | 
               | I would also like to be able to browse it online, but
               | this is a usability issue for strictly when I'm intending
               | to read it in a browser alone.
               | 
               | As to your final paragraphs referring to communication
               | and me "explainig to you that this stuff is normal", you
               | specifically said that "We don't want to normalize
               | unnecessary behavior" which implies that you do not think
               | it is already normalized. You're also implying that I
               | should have altered my interpretation of your words when
               | you said that I was missing your point, even though you
               | didn't say I was missing your point until the same reply.
               | 
               | In any case, I think I understand your POV regarding
               | archives, and I disagree.
        
               | godelski wrote:
               | > I would prefer that it is distributed as a zip. It
               | allows me to easily get the entire file, and hash it to
               | make sure it's the same file as other people are getting,
               | and have an archive of it.
               | 
               | I mean hosting it on any GitHub alternative makes this
               | possible too. We also get better archival because when
               | things change, we can see. Considering this says "Part 1"
               | I expect things to change. History tracking is better for
               | archival.
               | 
               | > you specifically said that "We don't want to normalize
               | unnecessary behavior" which implies that you do not think
               | it is already normalized.
               | 
               | That's not accurate. Here's a counter example "We don't
               | want to normalize clickbait headlines." Clickbait
               | headlines are already normalized, that does not mean we
               | want them to be nor does it mean we should accept them
               | and not fight against them. I'm sure you can find many
               | other similar examples.
        
           | andrewflnr wrote:
           | To me the question is, why would you put source code on
           | github if you're _not_ going to make it uncompressed? What 's
           | the point of using a source code hosting website if your
           | payload is a link to an upload site? Pastebin sites have been
           | around for years.
        
             | godelski wrote:
             | Exactly. Similar questions                 Why does a user
             | need to download a file to achieve the goals? Does doing so
             | provide added utility?       Does obscurification provide
             | some benefit?       Does distribution in this manner help
             | normalize environments which scammers take advantage of?
             | 
             | I'd argue:                 - Don't make users download
             | things they don't have to.       - Serving in plain text
             | gives higher utility as users can view it on any device
             | (e.g. mobile. Am I the only one that reads repos on
             | mobile?)         - A GitHub alternative also provides the
             | capacity to download an archived zip, thus achieving any
             | benefits that aren't obscurification related         - Git
             | helps for better archiving as we can have a track record of
             | commits and changes (this is labeled "Part 1"!)         -
             | Did no one else notice that there are ".github" directories
             | with workflows? But there is no ".git" folder? I'd honestly
             | like that...       - While a zip itself is not an
             | executable and not generally dangerous in of itself,
             | scammers (hackers) do take advantage of such environments.
             | Because you can... change a file extension. Or because a
             | user may double click the zip to extract, but this will
             | cause execution. Or idk, hackers are fucking smart and
             | people are dumb.
             | 
             | I'm a bit peeved that people feel the need to explain to me
             | that a zip isn't nefarious in of itself, because that's not
             | what I was concerned with (and that there's several such
             | comments and we don't need to keep repeating the same
             | comment...). My concern is with how such formatting is (as
             | best as I can tell) not necessary, suboptimal, and
             | normalizes practices that nefarious actors take advantage
             | of. This topic is obviously hot, so I won't be surprised if
             | there are "alternative links" that could just contain
             | straight up maleware. Yeah, the user has to execute it, but
             | people are dumb, lazy, and/or tired and there is a *
             | _better*_ form of distribution that just doesn 't leave
             | this script-kiddy style attack around. Like for fuck's
             | sake, people at intelligence agencies plug in USBs they
             | find on the ground...
        
         | wibbily wrote:
         | I've nothing to do with the leak itself, so can't help you
         | there. But did check - both archives are identical, and contain
         | a Node project that seems to match what is claimed. (Run it at
         | your own risk.)                 $ md5sum lam.zip
         | 3a78b14e1379ac5c059dbbe5660fca8a lam.zip
        
           | godelski wrote:
           | Thanks! I don't actually assume that the person is being
           | nefarious, but I think it is also important to make sure that
           | they understand these things. Especially if we're talking
           | about scams.
           | 
           | Scams take advantage of what is normalized, it is how they
           | fly under the radar and bypass people's bullshit detectors.
           | It's why a safety vest, hardhat, and a clipboard is the most
           | covert disguise around. So one of the best ways to prevent
           | scams is to normalize behavior that is harder to take
           | advantage of! (same reason people fall for fake voice scams,
           | because we're so used to distortion in calls anyways. A
           | glitch poor voice can be difficult to distinguish from poor
           | cell reception)
           | 
           | As for the filesizes, I assume it is just the websites
           | reporting incorrectly. Pixeldrain reports 188 MB compressed
           | and 510 MB uncompressed. Mega reports 179.r MB. Pixeldrain at
           | least shows all the files, which look to not have been
           | cleaned up since they have things like .DS_Store. But at
           | least the files are individually downloadable.
        
             | fwip wrote:
             | 179 MB if a megabyte is 1024^2 bytes, 188MB if it's 1000^2.
             | 
             | Zip files aren't evil, just unzip them and look inside.
        
               | godelski wrote:
               | I think you're missing the point of my comment.
               | 
               | The point of the comment and request is about not
               | requiring technical knowledge and minimizing amount of
               | necessary thinking. The point is about helping stop
               | scammers in the first place!
               | 
               | > 179 MB if a megabyte is 1024^2 bytes, 188MB if it's
               | 1000^2.
               | 
               | This is not entirely correct though because MB != MiB. Us
               | on HN will probably know this but proper labeling helps
               | prevent mistakes. The improper labeling requires us to
               | think more when considering security, which is bad
               | security (not that you shouldn't think, but I'm saying
               | "don't set off alarms when you don't need to set off
               | alarms")
        
               | dns_snek wrote:
               | > This is not entirely correct though because MB != MiB.
               | 
               | The point the parent was making is that the file is
               | 188026773 bytes long. One site represents that as 179 MB
               | (base 1024) and the other one as 188 MB (base 1000). Your
               | complaint is therefore with one of the websites and not
               | with the uploader.
        
         | kish_kush wrote:
         | uncompressed and compressed have nothing to do with what you
         | said. you can choose to run the code or not, but it doesn't
         | have to do with the uncompressed thing.
        
           | godelski wrote:
           | You're missing the point. The point is to make it harder for
           | scammers. Yes, I can safely extract files but on many systems
           | if you double click a zip instead then there you go. Either
           | way, it is always best to not download when you don't have
           | to.
           | 
           | The question here is "is there a reasonable alternative that
           | doesn't require the user to download." The answer is
           | unambiguously "yes" and unambiguously has higher utility.
        
         | indrora wrote:
         | a zip archive is not executable unless something has gone very
         | wrong.
        
           | godelski wrote:
           | You're missing the point. As I've been explaining in other
           | comments which have expressed the same thing as you have
           | (please read to reduce noise and repetition), the point is
           | about not normalizing environments which scammers can easily
           | take advantage of. And clearly, the request has higher
           | utility, so in either way, it is an advantage.
        
         | meindnoch wrote:
         | Bro, it's a zip file. It won't set your computer on fire.
        
           | godelski wrote:
           | Bro, no one claimed this.
        
         | mrstone wrote:
         | At the very least, here's a scan from Jotti.
         | 
         | https://virusscan.jotti.org/en-US/filescanjob/svl9focwgt
        
       | extr wrote:
       | I really think this + the humane AI pin would be super
       | interesting products if they made them hackable. The hardware is
       | super cool, no problem if the software isn't there yet, it's not
       | like they're being sold at Best Buy, I'm not worried about my mom
       | acquiring one of these and getting her passwords leaked. Who
       | cares if the auth flow is super hacky/insecure? Let us self host
       | it! Let the community create more playwright scripts!
        
         | roughly wrote:
         | Humane is founded and populated by ex-Apple folks, so I
         | wouldn't hold my breath.
         | 
         | You're right, though - it's a bit weird because ostensibly the
         | interface is via the GPT system and you've gotta work through
         | those interactions (same problem Alexa had), but given how
         | early both the platform and the product category are, they'd
         | benefit big from letting early adopters build capabilities for
         | them.
        
           | harryp_peng wrote:
           | Worst possible thing is that bozos learn the closed source
           | model. The closed source model only worked because they had
           | the Great SJ.
        
         | 999900000999 wrote:
         | Agreed.
         | 
         | The Rabbit R1 looks like it would be the perfect device to play
         | with.
         | 
         | As is I just assumed Rabbit was sending off pictures and stuff
         | to a Chat GPT API or something. I never assumed the models ran
         | on device
        
         | disconcision wrote:
         | this exists!
         | 
         | https://www.openinterpreter.com/
         | 
         | even their hardware is (apparently) open-sourced
        
           | spaceship__sun wrote:
           | But how is an hardware version of a mic/speaker anything
           | interesting? The humane AI pin, to which you wear and gets
           | contextual data, is the only wearable of interest to hackers.
        
         | vineyardmike wrote:
         | If you want something fun and hackable look at these glasses:
         | 
         | https://brilliant.xyz/products/frame
         | 
         | They have shipped products before, and they include a bunch of
         | code - _today_ on GitHub to start hacking with.
        
       | phh wrote:
       | > In reality, they're simply relying on several Playwright
       | automation scripts to do the job for you, which is why they only
       | support four apps: Spotify, Midjourney, Doordash, and UberEats.
       | 
       | I think that part is mostly fine? I'd rather make give a LLM
       | access to https://woob.tech to be my personal assistant while
       | parsing 99% less noise, than have a LLM that parse and understand
       | stupidly complicated web pages, and randomly fail at the task
       | because the name of my doctor is bobby drop tables.
       | 
       | That being said, it can be interesting to use LLMs to assist
       | creating woob plugins.
        
         | vunderba wrote:
         | Midjourney does not have a public API and I'm pretty sure that
         | automating a Midjourney account is against the TOS, so I
         | wouldn't expect that functionality to last long.
        
         | saltsaman wrote:
         | The problem is that they claim to have developed a
         | groundbreaking Large Action Model when in fact it's just a
         | playwright wrapper
        
           | ach9l wrote:
           | you can't automate playwright without a decision making
           | component in front of it, they are definitely using a
           | transformer there. one could train a llama and make it
           | perform triggers to playwright automations. you can even get
           | deep into transformer tokenization and create action tokens
           | and a formal grammar for your generation, build a parser on
           | top of your predict function and have a "lam" working. the
           | fact that they use playwright does not imply it is not
           | generative ai. i'd say it is really hard to do those actions
           | without a transformer involved
        
       | Jonovono wrote:
       | What NFT projects were they involved with?
        
         | ugh123 wrote:
         | /s?
         | 
         | Underrated comment.
        
           | itishappy wrote:
           | > Sadly, this shouldn't come as a shock to anyone who's done
           | minimal due diligence on the team. After all, they were still
           | hawking NFTs just two years ago.
        
           | tripletao wrote:
           | No sarcasm,
           | 
           | https://web.archive.org/web/20221203132009/https://gama.io/
        
       | ipsum2 wrote:
       | I implemented my own DIY version of Rabbit at a hackathon using
       | Playwright and VNC. I feel extremely validated that they use the
       | same things that I thought of.
        
       | ugh123 wrote:
       | If it can do what it claims to do, which is automate on top of
       | existing apps by your voice, whats the difference? Seems
       | innovative regardless of the tech underneath.
        
       | amiantos wrote:
       | I'm really into AI stuff but both the AI Pin and Rabbit R1
       | underwhelm. They are products that don't need to exist if the
       | problem they're solving was truly solvable right now, because the
       | best place for that problem to be solved is already in our hands:
       | our phones. But we're not all talking to AI assistants in our
       | phones all day. Why? Because the technology isn't good enough to
       | do it yet? Because people don't want to talk to digital
       | assistants? Once the tech is good enough that it can motivate
       | ordinary people to look silly talking to their phones outloud,
       | it'll be on our phones and easy to use, and there will be no need
       | for kitschy little handheld devices. No one wants to carry around
       | another device.
       | 
       | That said, smart glasses sound like a great idea to me, but I
       | wear glasses all ay long, so I am extremely biased. I don't think
       | most people want to voluntarily wear glasses to just put a
       | computer on their face, so I wouldn't bet on glasses, either.
       | Sorry, Zuck.
        
         | azinman2 wrote:
         | I don't know why I need to keep saying this, but the point of
         | the pin is to replace phones. It's meant for you to have access
         | to digital services while staying in the moment and avoiding a
         | screen. Many ppl are addicted to their smartphones, so they
         | propose something different.
         | 
         | You can like and prefer a phone but it's their raison d'etre.
        
         | imiric wrote:
         | It wasn't that long ago that it wasn't socially acceptable to
         | have phone conversations in public wearing Bluetooth ear
         | pieces, whereas now we don't think twice about it. A few
         | decades before that, the same thing with portable music players
         | and headphones.
         | 
         | Society adapts quickly to technology, but, as you say, the tech
         | needs to be good first.
         | 
         | I think we've reached that point with voice recognition and AI
         | assistants. It's now a matter of time until someone connects
         | the pieces into a functional and accessible product.
         | 
         | The reason smartphones are not the devices to get us there is
         | because they're not a good fit for this use case. Pulling out a
         | rectangular slab with a huge screen out of your pocket every
         | time you want to interact with a voice assistant is enough of a
         | UX hurdle that most people won't do it, even if it would be
         | socially acceptable. Even if this was in a watch form factor,
         | which we'll surely see as well, just bringing your arm close to
         | your face would get slightly annoying over time.
         | 
         | So a light pebble device you can pin on your shirt or wear as a
         | necklace seems like a good form factor for this. The Limitless
         | Pendant is another recent contender, and seems like a better
         | thought out product compared to the Humane Pin. These devices
         | aim to be unobtrusive, and disappear into the background, yet
         | still remain deeply integrated into our lives. This is what
         | technology is trending towards. I reckon the smartphones of
         | today will seem primitive in a few decades, replaced by
         | seamless VR/AR in glasses and primarily voice-driven wearable
         | tech. We're currently in this transitional period where
         | companies are investing in high-risk products to see what
         | sticks, but eventually someone will launch something that
         | resonates. Just like Apple did for smartphones in 2007.
        
           | mulderc wrote:
           | Why wouldn't the device just be your ear buds connected to
           | your phone?
           | 
           | I already use siri all the time as I usually have my AirPods
           | in and it works great. That seems like a much more likely
           | device to access our digital assistance than some necklace or
           | pin. Better yet, you could just pair it with your smartwatch
           | and not even need the phone.
           | 
           | I'm still deeply skeptical on voice driven tech as we have
           | had that available and easy to use from various devices for
           | over a decade now and it hasn't taken off for tons of
           | reasons. I just am not going to have a conversation with my
           | computer with others around.
        
         | vineyardmike wrote:
         | So I agree generally,
         | 
         | BUT existing phone companies have an incentive to maintain the
         | app-centric world that keeps their app stores profitable, and
         | app companies have an incentive to lock you into their app to
         | keep customer loyalty and be "more than an API".
         | 
         | All that's to say, the Rabbit idea of manually scripting
         | against apps to allow "business as usual" for all these
         | individual parties who wouldn't want to collaborate fills a
         | void that existing players don't have incentives to fill.
        
       | eterpstra wrote:
       | Maybe they hacked together something that can feasibly me
       | marketed as an AI-assistant knowing that whatever they build now
       | will get "steamrolled" by GPT-5 (Sam's words, not mine). When
       | GPT-5 gets released, update the OS and it'll work as
       | advertised... EZ-PZ!
        
       | xori wrote:
       | Not much here explicitly in the source code dump. A little
       | insight into their worker node infra but no "secret sauce" imo.
        
         | saltsaman wrote:
         | Isn't the secret sauce just VNC with playwright? What more do
         | you need to achieve 80% of what they are showcasing (basic
         | doordash orders, spotify controls)?
        
           | rhinoceraptor wrote:
           | I can't find any purported auomation scripts for those
           | services as claimed in the Github page. There is a reference
           | to "cm-spotify-client" which seems to be some sort of custom
           | integration code they've written, but other than that there
           | is no reference to doordash, midjourney, or uber eats. This
           | dump seems to just be the code/infrastructure to run
           | chromium/playwright in kubernetes, wrapped in a Node API to
           | accept commands, persist/hydrate browser state, etc.
        
       | mafuyu wrote:
       | Pretty much in-line with my expectations. I ordered one because I
       | thought the design was neat, and I was interested in hacking
       | around and flashing my own stuff onto it. The pricing was clearly
       | at or below cost.
       | 
       | Looking at just the concept (and ignoring execution), I don't
       | really see the point of this thing? The whole thing is a feature
       | that could exist on a smartphone. The dream of an AI agent that
       | you can converse with to replace your smartphone could be
       | compelling, but nowhere close to reality yet. Even then, the big
       | smartphone OS companies are obviously better positioned for this.
       | The smartphone is the hub for all your information, plus they
       | have years of voice assistant, automation, and home IoT
       | integration to build off of.
       | 
       | Humane was silly because it was a smartwatch without any of the
       | proper software support, but Rabbit is essentially doing the same
       | but targeting a smartphone replacement. If you really want to
       | break out and try to dethrone smartphone vendors, you'll have to
       | come up with something more compelling than a worse user
       | interface to a poorly made software platform. That's a software
       | feature you're building.
       | 
       | In some sense, I do think Rabbit had a better approach than
       | Humane, though. Getting a bunch of low-priced "toy" devices into
       | the market that are just a frontend to your server software could
       | get you off the ground. The software needs to exist, though...
        
       | animex wrote:
       | I just couldn't fathom the big three phone platforms not
       | implementing this on a device that we all have and is capable of
       | same if-not better dynamic voice integration.
       | 
       | At the very least, I hope products like the Rabbit spur these
       | companies to start innovating again. Even if they are smoke &
       | mirrors, the interest shows there's demand for these features.
       | 
       | Site Note: I've noticed Google Home's voice assistant has
       | declined over time -- it used to handle complex queries and now
       | it can barely understand simple directions. It used to understand
       | me perfectly in the noisiest environments and now it makes many
       | transcribing errors.
        
         | redserk wrote:
         | Like in iOS with SiriKit?
         | 
         | Blame app developers for prioritizing implementing less useful
         | features.
         | 
         | https://developer.apple.com/documentation/sirikit/
        
           | Wowfunhappy wrote:
           | The problem with Siri is their voice recognition sucks,
           | especially compared to e.g. Whisper.
        
       | iamleppert wrote:
       | Who cares how its made if you can make a bag from it before
       | anyone is the wiser? The point in all this stuff is to make bags
       | of money, whatever way you can do that don't matter as long as
       | you gettin the bread.
        
       | vunderba wrote:
       | One of the bizarre talking points in defense of the existence of
       | rabbit was to get away from our phones. It's just completely
       | inexplicable to me because it's not like it was ever intended to
       | be a replacement, the only difference is congratulations you now
       | have to lug around two separate brick shaped appliances wherever
       | you go...
        
       | stranded22 wrote:
       | They gave 12 months perplexity pro with it - and I am already a
       | subscriber. So, I basically paid a bit extra for another 12
       | months and a rabbit r1 to play with.
       | 
       | If it doesn't work how I want, I should be able to sell on whilst
       | keeping the perplexity pro sub.
        
       | anon115 wrote:
       | LOL
        
       | m3kw9 wrote:
       | The UIUX isn't good, AI models are useless without good user
       | experiences
        
       | serf wrote:
       | I never understood the appeal outside the cute form-factor, all
       | of the demos were absolutely terrible.
       | 
       | an aside : npr doesn't like the 'spade' comment, although I think
       | the explanation is kind of iffy.[0]
       | 
       | [0]:
       | https://www.npr.org/sections/codeswitch/2013/09/19/224183763...
        
       | latentcall wrote:
       | I thought the idea of the Rabbit R1 was cool. I have a strong
       | feeling this will end up like the Humane Pin, which is sad. I'm
       | glad companies are trying something different.
       | 
       | I'd love to be able to use my phone hands free without having to
       | look at it, and interface with ChatGPT/Claude/whatever but I am
       | not sure if it's possible? Siri works very poorly and is
       | unreliable. I'd like to be able to use an LLM as a personal
       | assistant. Set timers, call people, message people, but also be
       | able to ask questions like the voice chat function in the ChatGPT
       | app. Maybe one day!
        
         | mulderc wrote:
         | I must be a Siri unicorn but for setting timers, calling
         | people, messaging people, controlling my smart home, adding
         | things to my shared shopping list, adding items to my 3rd party
         | task manager, controlling my music. It works great!
        
       | NetOpWibby wrote:
       | According to their CTO in the Discord[0]:
       | 
       | > If someone spends enough time with the login minions they can
       | extract these code. But these code are locked down and are
       | sanitized. LAM lives elsewhere. This is someone looking at the
       | rabbit hole not understanding how it works. And tries to be
       | smart.
       | 
       | [0]:
       | https://cdn.discordapp.com/attachments/1185274946981732374/1...
        
         | patleeman wrote:
         | Can somebody translate this?
        
           | floren wrote:
           | "Shit, shit, shit, shit! Dissemble!"
        
           | _heimdall wrote:
           | The original claim in the code dump is that no ML tools are
           | used at all and the tool is just leaning on Playwright to
           | automate specific actions on a website.
           | 
           | The CEO here is claiming that the ML code is being run
           | outside this code base and that the original claim is being
           | made by someone who doesn't know how the code works.
           | 
           | The CEO's mention of sanitized code isn't as clear to me,
           | that can mean different things. Compiled code can be
           | considered sanitized since it likely isn't human readable,
           | obfuscated code makes that harder, and removing some code all
           | together would be the most effective. The problem with
           | removing code all together is that you would still find code
           | paths that just can't be executed at all, leaving some trail
           | of what code was removed. That wouldn't leak any secrets
           | obviously, but would support the argument that code has been
           | removed and the codebase is being misread.
        
           | threeseed wrote:
           | I think it means to say that:
           | 
           | 1) The got the code by bruteforcing the login credentials on
           | device.
           | 
           | 2) Server-side code is not accessible which is where the LAM
           | runs.
        
       ___________________________________________________________________
       (page generated 2024-04-23 23:01 UTC)