# Replit Drops Anthropic, Amazon’s Bee AI & Tesla’s “Flying” Roadster

An episode of [AI Chat: AI News on Claude, ChatGPT, Gemini, Grok, Perplexity](https://invisiblenetwork.ai/podcast/1669990482), 17 Oct 2026, 15 min.

Audio: https://content.rss.com/episodes/365073/3241369/ai-chat-podcast/2026_10_17_00_24_42_3eea03b4-5fea-45c0-badd-470ed31b3719.mp3

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! In this episode, we cover Replit's decision to discontinue its partnership with Anthropic and the launch of Amazon's new Bee personal AI devices. Additionally, we explore Tesla's recent demonstration of its new Roadster and clarify why this vehicle is not designed for self-driving capabilities. Show Links • Get the top 80+ AI Models for $8.99 at AI Box: https://aibox.ai • How I Grow and Scale My Business with AI: https://www.skool.com/aihustle • Get the AI Chat Daily Newsletter: https://www.aichatdaily.com/newsletter

## Transcript

[0:00] Welcome to the podcast. Replit says that they
[0:02] are no longer working with Anthropic. We have
[0:05] a brand new self -driving or I guess the lack
[0:09] of a self -driving car coming out of Tesla with
[0:11] their new flying roadster. And I think there's
[0:14] actually some interesting implications for AI
[0:16] and self -driving cars that Elon said while unveiling
[0:19] all of this. Jev is apparently Vercel's fastest
[0:23] adopted AI model. However, I'm going to explain
[0:25] a little bit about what Jeff does and why that's
[0:27] a big deal. Amazon is introducing two new AI
[0:31] products from Bee. They have Bee Frontier and
[0:33] Bee Eclipse. I've hosted the Bee devices in the
[0:36] past. I'm going to tell you a little bit about
[0:38] the new Bee devices, how they work and what they
[0:40] do. And then we have the Fed chair, Kevin Warsh,
[0:44] that is warning everyone about an AI... shock
[0:47] for the economy. And he says that he is promising
[0:50] that he's going to continue battling inflation
[0:53] despite that. And then we have a couple of funny
[0:55] memes over on X we're going to get into. But
[0:56] before we do, I wanted to mention, if you want
[0:59] to get access to over a hundred different AI
[1:01] models inside of ChatGPT or inside of Clot or
[1:05] inside of Gemini or wherever you, whatever your
[1:07] main workstation is for AI. I would love for
[1:10] you to check out the AI Box MCP. It's a connector
[1:13] that basically brings in all the other AI models
[1:15] wherever you work. So with something like Claude,
[1:17] you get access to image generation from ChatGPT.
[1:20] You get access to video generation from Google.
[1:22] You get access to audio generation from 11 Labs
[1:24] or vice versa. In all of those environments,
[1:27] you get access to all of the different AI models.
[1:29] So 100 different AI models. It starts at $8 .99
[1:32] a month. And I think we have even discounts to
[1:34] make it cheaper for annual plans. If you need
[1:36] more tokens, we do that. I recently... had a
[1:39] project where i was able to generate over 1200
[1:41] images for a website and saw the traffic explode
[1:45] because these were images that were assets for
[1:48] something in particular so anyways if you're
[1:50] interested in getting access to all of these
[1:52] different tools and being able to chat with 100
[1:55] different AI models in one place. Go check out
[1:57] AIbox .ai. I'll leave a link to it in the description.
[2:00] All right, we're going to kick off with a little
[2:02] bit of drama here. And I don't know if this is
[2:04] real drama, but basically the CEO of Replit said
[2:07] that they're no longer working with Anthropic.
[2:09] And he said it comes down to a couple different
[2:11] things. This is the quote from him. He said,
[2:13] I think ChatGPT is better now. He said that they
[2:17] are using internal benchmarks to help assess
[2:18] metrics like performance. He said, we publish
[2:21] a public benchmark as well. and consistently
[2:23] Sol and Astra perform better on public or on
[2:26] building vibe -coded applications from scratch.
[2:28] It was a decision based on both costs, partnership,
[2:31] and performance. And for all of these reasons,
[2:33] we just felt like they were a better partner
[2:35] for now. I'm going to be honest, as someone that
[2:38] has both the Mac subscriptions for Cloud and
[2:41] for ChatGPT, I find myself using ChatGPT much
[2:43] more right now. Probably for the last six months,
[2:45] it was Cloud before that. And I think especially
[2:48] once you start using things like Cowork, ChatGPT
[2:51] Cowork. or Claude Cowork, I'm finding that Cowork
[2:54] or Chachapiti Work is better than Claude's offering
[2:57] there. And I know there's a bunch of different
[2:59] tools. We got Muse, we got Grokbots, we got I'm
[3:02] sure Google's working on something. But right
[3:04] now, Chachapiti is the winner. So I'm not surprised
[3:07] by all of this. There is some, you know, some
[3:10] blog posts from September that were saying Anthropix.
[3:13] Fable 5 and Claude Fable 5 .1 were helping Replitz
[3:17] agents run. So I don't think they're like completely
[3:20] detached from them. But I think he's just making
[3:22] a statement saying, look, this is the main model
[3:24] that we're using to power a lot of what's going
[3:26] on. I don't think this is means they're going
[3:28] to, you know, they're completely cutting it off.
[3:29] As far as the growth of his company, one thing
[3:31] that's interesting, if you've been following
[3:33] Replit for a long time, he said that they're
[3:34] not sharing their ARR numbers right now for growth.
[3:37] But he said that enterprise was 2 % of revenue
[3:39] last year. And this year is about 15 % of revenue.
[3:43] And then it's more than 50 % of the revenue total
[3:46] is coming from B2B. So as a company, I think
[3:49] they're growing a lot and seeing a lot of value
[3:51] from all of the vibe coding that's going on right
[3:54] now. And it's interesting that... is what is
[3:56] powering that at the moment. OK, I just had to
[3:58] share a meme that I saw. If you're watching on
[4:01] video on Applecast, you can see video now on
[4:03] Spotify, of course. But there's a there's a meme
[4:05] because yesterday Elon Musk unveiled the new
[4:07] self not self -driving roadster. He said this
[4:10] is one of the very last vehicles that Tesla will
[4:12] ever make. That is not. going to be self -driving.
[4:16] And the reason is because this vehicle flies.
[4:18] The new Roadster has rocket thrusters on the
[4:21] bottom, which when I first saw, I was like, oh,
[4:23] no, this is going to like burn holes in the road.
[4:25] It's like a rocket. But it's actually just compressed
[4:27] air tanks. And from this compressed air tank,
[4:29] the entire car can shoot in the sky, fly forward,
[4:33] land, do flips. It's very wild. This is it's
[4:37] the craziest thing I've ever seen. But anyways,
[4:39] the meme right now is that someone said traffic
[4:41] is a bit heavy on the X. are on the z -axis today
[4:45] and it just shows like a street with like a hundred
[4:47] different cars that are all flying and they're
[4:49] all flipping at different angles and driving
[4:51] all crazy i mean it's it's gonna be pretty wild
[4:53] if this becomes a thing where you're actually
[4:55] flying to work or flying to wherever you're going
[4:58] and yeah traffic in the sky is gonna be different
[5:01] than traffic on the ground i have no idea how
[5:05] we would how we're gonna deal with millions of
[5:07] cars flying through the air but I do think it's
[5:10] interesting that it brought up a bigger point.
[5:12] The reason I brought this up on the AI podcast
[5:16] today is because he brings up an interesting
[5:19] point, which is this is the last car that they're
[5:20] going to make that's not self -driving. Meaning
[5:22] they're not training an AI model to fly your
[5:26] car around because there's no historical context.
[5:29] How did they train self -driving cars? They stuck
[5:31] cameras on real cars. And as the real cars drove
[5:34] around, it gathers data and teaches the car how
[5:36] to drive. This is how humans drive. This is how
[5:38] to drive. There's not a lot of precedent for
[5:40] how a human might drive a rocket -propelled car
[5:43] through the air. And so they're not able to train
[5:46] a model on something that there's no data for.
[5:49] So anyways, I just thought this was fascinating.
[5:50] It made me wonder. What other areas are we not
[5:53] going to be able to train robots or cars or other
[5:57] AI to do things just because we don't have a
[6:01] lot of data? I mean, they're completely new industries
[6:03] that are being invented right now. So it's an
[6:05] interesting thought experiment. All right, Jev
[6:07] is apparently Vercel's fastest adopted AI model
[6:10] ever, according to, well, Vercel, but there's
[6:13] a whole report in TechCrunch about this. Vercel
[6:16] published a bunch of data that shows that within
[6:18] two weeks of launching Jev, which is, it's like
[6:21] a decision model, it was being used by 35 % of
[6:24] paid teams and nearly double the reach of the
[6:27] second most popular model, which was Sonnet 5
[6:29] .5. They're giving Anthropic a bit of a run for
[6:32] their money. They said it's the fastest adoption
[6:34] model has ever seen in the first two weeks. According
[6:37] to them, they published this as a limited release
[6:40] on September 15th. So this wasn't something that
[6:42] everyone automatically got access to. Only some
[6:44] people got access to it. It's basically... Designed
[6:47] as a lightweight tool for choosing between a
[6:49] few predetermined outputs. And it is limiting
[6:52] the outputs. And by doing that, it's letting
[6:54] the model run cheaper and faster than alternatives
[6:56] like Sonnet. So it makes it really good for classifying
[7:00] huge amounts of data quickly. It's a new approach
[7:03] and it has gotten a lot of buzz. A lot of people
[7:05] are talking about this. It has inspired Amazon
[7:08] both to release competitors to it. And I think
[7:11] there's a few others that are working on some
[7:13] as well. B is introducing, or I guess Amazon
[7:17] is introducing, to do AI products from B, the
[7:19] company that they purchased, which TLDR, I've
[7:22] talked about this before. I don't want to just
[7:23] roast them forever, but it's basically like a
[7:25] plastic microphone that would strap as on a wristband
[7:28] to your hand. I purchased her for 50 bucks to
[7:31] test it out. Did not find it very interesting
[7:33] or good. They have created some new products
[7:36] that I do think are interesting. You will need
[7:38] Alexa Plus in order to use them. Two products.
[7:41] The first one that I think is interesting is
[7:43] basically like a magnet. And it looks basically
[7:46] like those chargeable or those like phone batteries,
[7:49] like those external phone batteries you snap
[7:50] onto the back of your iPhone with the MagSafe.
[7:52] It's basically one of those, except. That's the
[7:55] device and it can snap onto the back of your
[7:57] phone and it can be always recording. And the
[7:59] concept, they also have a bracelet that is the
[8:01] same thing. And the concept between the bracelet
[8:03] and this device is basically it's always recording
[8:07] and transcribing everything you're doing. It's
[8:09] secure. It's on device. It doesn't have to, you
[8:12] know, and it automatically deletes it after a
[8:14] certain amount of time. But it reminds you of
[8:15] things. And in the specific demos that they've
[8:17] talked about. They say, have you ever discussed
[8:21] plans with a friend only to immediately forget
[8:23] the date and time you decided on? It's a relatable
[8:26] scenario. Our days are full of information that
[8:28] matters, but holding on to all of it while staying
[8:30] present can feel impossible. V is a personal
[8:32] AI that was created to solve that. It takes notes
[8:35] of the details you might otherwise forget and
[8:37] turns them into useful conversations and summaries
[8:40] and to -dos, reminders. and inside so i mean
[8:44] i get the idea here a lot of people will criticize
[8:46] this saying like well what's the difference between
[8:48] just like i don't know having an app on your
[8:50] phone that does this i guess having an app that's
[8:52] always running and would probably suck your battery
[8:54] dry maybe something like this you can make an
[8:57] argument that they could do it with less battery
[8:59] maybe it's more secure you're not worried about
[9:02] where the data is going transcribing everything
[9:04] maybe the app will crash at a certain point i'm
[9:06] not sure there's just probably some arguments
[9:08] but basically you could do this with an app on
[9:10] your phone theoretically so i think it's interesting
[9:12] i'll be curious to see if amazon's able to make
[9:14] this something that is popular we have like a
[9:17] lot of people have suggested making stuff like
[9:19] this i think the friendship pendant was similar
[9:21] to this i think it just feels like a lot to have
[9:24] a whole credit card or battery puck thing strapped
[9:27] to the back of your phone always just to record
[9:30] so i'm curious to see if if a lot of people use
[9:33] that although i will say Like kudos to them for
[9:36] coming up with something unique. This is the
[9:37] first time I've seen that idea of snapping something
[9:39] on the back of your phone. I've seen the bracelets.
[9:41] I've seen necklaces. We've seen the Meta Ray
[9:43] bands. We've seen, you know, glasses, necklaces,
[9:45] watches, probably rings will be next if they're
[9:48] not already out there. And I just missed announcement,
[9:50] but it's basically things that you already would
[9:53] wear. And now we're adding microphones into them.
[9:55] And phone seems to be the snap on the back. Your
[9:58] phone seems to be the next one. I know for people
[10:00] that. value privacy this is probably an absolute
[10:02] nightmare I don't think my wife would ever strap
[10:05] a microphone to herself intentionally but hey
[10:08] maybe some people would they say no audio is
[10:10] ever stored audio is processed into insights
[10:12] in real time and deleted after processing this
[10:15] and no one else can access it although not to
[10:17] be a wet blanket but I remember when people accidentally
[10:20] got access to other people's Amazon Alexa voice
[10:23] recordings that's embarrassing for them It said
[10:25] you're always in control. You get personalized
[10:27] Alexa access. You decide what to keep, yada,
[10:30] yada. So I'll be curious to see how well these
[10:32] sell. The Bee Eclipse is going to be available
[10:35] for $100. And the Bee Frontier is going to be
[10:39] available for $74. Too many products out of Amazon.
[10:42] Let's see how well these things sell. The federal
[10:45] chair, Kevin Warsh's warning of an AI shock.
[10:48] I talked about this a little bit yesterday. And
[10:50] I mean, the concept is these AI stocks have pushed
[10:53] the market up so high. You know, a lot of people
[10:55] say we would have entered a recession in the
[10:57] last few years if it wasn't for AI and the way
[11:00] that that's pushed the economy. In all of this,
[11:02] he said, quote, if you're a central banker, you're
[11:05] in the business of trying to ensure stable prices
[11:07] and full employment. You've got to recognize
[11:09] this as a shock. He was in Bangkok, Thailand
[11:12] at a conference. Apparently a huge... huge surge
[11:14] in AI spending has helped boost the US economy
[11:17] in recent quarters. I mean, we know every major
[11:20] hyperscaler is spending aggressively on data
[11:22] centers and aggressively on training data and
[11:24] everything else required. So in theory, all of
[11:27] this investment and all of this activity for
[11:30] chips manufacturing data centers could put upward
[11:32] pressure in general on consumer prices unless
[11:34] demand is offset by a new supply of goods and
[11:37] service. He said, quote, it's obviously hitting
[11:40] the demand side of the economy before it's hitting
[11:42] the supply side of the economy. He said the path
[11:45] from where we are to there is likely to be a
[11:48] very bumpy one. When he was talking, he said
[11:50] he's really optimistic about AI, but said it's
[11:52] getting pretty bumpy. He said, we really don't
[11:54] know what the market structure will be of these
[11:56] new technologies. Who's going to be making the
[11:59] money? So anyways, I think there's a lot of debates
[12:03] even happening at the Fed right now. The last
[12:06] quote I'll leave you with from him is that he
[12:07] said, it's hard to be in discussions with central
[12:10] bankers and not compare notes and hear the same
[12:12] things about the surge in prices. I've heard
[12:14] resolve from my colleagues around the world to
[12:16] deliver price stability. I've certainly delivered
[12:18] that resolve. So inflation is happening all around
[12:22] the world. And I think specifically, you know,
[12:24] like he was mentioning, this is a global phenomenon.
[12:27] And he believes perhaps AI is helping drive up
[12:30] the price, right? Energy is in more demand. There's
[12:32] a lot of things that are in more demand. But
[12:34] at the same time, it's also creating you know
[12:36] a ton of jobs and a lot of money is going into
[12:38] the economy from this so it depends on how you
[12:41] look at that but his point is it might cause
[12:43] a surge in prices before we get all of the benefits
[12:47] of then the job creation and the revenue back
[12:50] into the economy, etc. So it'll be interesting
[12:52] to see what happens there. All right, we're going
[12:54] to end the podcast today with a really funny,
[12:57] I don't know if this is a meme, but like a funny
[12:59] moment I saw over on X. Apparently there's this
[13:01] clip where it says a Chinese bride gifts her
[13:04] partner flagship GPU at a wedding. He breaks
[13:06] down in tears. She hands, she gave him basically
[13:09] an Asus ROG Astral RTX. 50 -90 GPU. Those things
[13:13] cost about 12 grand in China. And yeah, basically
[13:16] she is like their wedding. She's in her wedding
[13:18] dress. She gives him this black bag and he's
[13:20] like opening it. He's in tears. Out of a black
[13:25] bag. And he goes and walks around. The meme is
[13:27] that he's like showing this GPU to everyone at
[13:29] the wedding. He's like it's Simba in the Lion
[13:32] King. He's holding that. He's really excited.
[13:34] All sorts of comments on this. Some people are
[13:36] saying what has happened to the modern man. Other
[13:39] people saying. go find yourself a wife that will
[13:42] give you 5090. I mean, these things are really
[13:45] expensive. So it's a it's a nice present. If
[13:48] you put on a credit card for getting married,
[13:50] then I guess you'll have to pay it off anyways.
[13:52] But for now, I think he's going to enjoy that
[13:54] GPU. And that's pretty, pretty funny clip. All
[13:57] right, guys, thank you so much for tuning into
[13:58] the podcast. Make sure you go check out AI box
[14:00] .ai if you want to get access to over 100 different
[14:03] AI models, we did a huge update last week, where
[14:07] we gave you the ability to make groups or favorites
[14:09] inside of AI box. So what I'm doing is I'm making
[14:11] a group with Gemini, Claude, ChatGPT, and Grok.
[14:15] I ask one question and all four of those models.
[14:18] type out their response at the same time, and
[14:20] I basically can scan between them and decide
[14:22] which model is best at answering what thing.
[14:24] I can compare them all side by side. Very, very
[14:26] useful. You can also do that with images, different
[14:29] image models, or audio or video. So if you want
[14:31] to check it out, go check out AIbox .ai, and
[14:34] I will see you guys all in the next episode.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.