AI Investing: for the AI Investor

Speakers analyze Anthropic's Claude Fable 5.2 performance, pricing, and Mythos 5.2 security features

11 min Listen to the original Episode page
Artificial IntelligenceLarge Language ModelsCloud ComputingCybersecuritySoftware Development

Summary

Speaker 1 and Speaker 2 discuss the release of Claude Fable 5.2, noting a 30% speed increase and 20% price cut. They evaluate the economic impact on heavy users and workflow automation, with Speaker 2 considering an upgrade to the $100 monthly plan for video editing tasks. Benchmark results show Fable 5.2 outperforming competitors like GPT-6 Astra and Gemini 4 Carbon on coding tests. The episode also covers the unreleased Mythos 5.2, highlighting its superior security capabilities and restricted access for vetted cybersecurity firms.

Episode sections

  1. 0:00 – 2:50

    Sponsor Reads and Intro

    The host reads advertisements for Staples Print and promotes a paid podcast subscription on Spotify/Apple called 'AI Chat' featuring an interview with the CEO of Cohere.

  2. 2:50 – 7:10

    Claude Fable 5.2 Performance and Pricing

    Speaker 1 discusses the 30% speed increase and 20% price cut for Claude Fable 5.2, noting its impact on long-running automation workflows and token usage efficiency.

  3. 7:10 – 8:07

    User Workflow Implications

    Speaker 2 considers upgrading from the $20 to the $100 monthly plan to use Claude for video editing, while Speaker 1 highlights the economic value of the price cut for heavy users.

  4. 8:07 – 11:20

    Market Comparison and Mythos 5.2

    Speaker 1 compares Fable 5.2's coding benchmarks against competitors like GPT-6 Astra and Gemini 4 Carbon, and discusses the restricted access and superior security capabilities of the unreleased Mythos 5.2.

Claims worth verifying

Cohere has raised over a billion dollars for their AI model.

Unidentified speaker (UNKNOWN) ·0:31

How to verify: Check Cohere's official press releases or financial news archives for funding announcements around the time of the podcast release (Nov 2026). For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Claude Fable 5.2 is approximately 30% faster than previous versions.

Unidentified speaker (speaker_1) ·2:50

How to verify: Review Anthropic's official blog post or documentation for the Fable 5.2 release notes regarding performance metrics. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

The price of Claude Fable was cut by 20%.

Unidentified speaker (speaker_1) ·1:10

How to verify: Check Anthropic's pricing page or official announcement for the cost per token or subscription tier changes. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Fable 5.2 is included in the Pro subscription for up to half of usage.

Unidentified speaker (UNKNOWN) ·3:44

How to verify: Review Anthropic's subscription terms and conditions for the 'Pro' tier to verify model access limits. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Speaker 1 has a $20 weekly limit on their Claude subscription.

Unidentified speaker (speaker_1) ·6:04

How to verify: Check Anthropic's current pricing tiers; note that Anthropic typically charges monthly, so this may refer to a specific budget cap set by the user or a legacy plan. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Speaker 1 spent $600 a month on Claude and ChatGPT subscriptions.

Unidentified speaker (speaker_1) ·6:24

How to verify: This is a personal anecdote; verify via the speaker's social media or blog if they publicly share financials, otherwise treat as unverified personal claim. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Fable 5.2 scored 49% on the Frontier Code 2.0 benchmark.

Unidentified speaker (speaker_1) ·9:07

How to verify: Search for the 'Frontier Code 2.0' benchmark results published by Anthropic or the benchmarking organization. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Opus 5.5 scored 45.5% on the Frontier Code 2.0 benchmark.

Unidentified speaker (speaker_1) ·9:27

How to verify: Search for historical benchmark data for Opus 5.5 on the Frontier Code 2.0 leaderboard. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

GPT-6 Astra scored 42% on the Frontier Code 2.0 benchmark.

Unidentified speaker (speaker_1) ·9:27

How to verify: Search for OpenAI's GPT-6 Astra performance on the Frontier Code 2.0 benchmark results. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Mythos 5.2 scored 89% on the CDE rediscovery benchmark.

Unidentified speaker (speaker_1) ·10:07

How to verify: Search for Anthropic's official announcement or technical report detailing Mythos 5.2's performance on the CDE rediscovery benchmark. For recent events, verify 2026-08-15 through 2026-11-15. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify the 'Frontier Code 2.0' benchmark methodology and dataset to confirm the 49% score and competitor comparisons.
  • Check Anthropic's official pricing page and API documentation for the stated 20% price cut and 30% speed increase claims.
  • Review the 'CDE rediscovery benchmark' details to understand how Mythos 5.2's 89% score and 14% vulnerability improvement were measured.
  • Investigate reports on the black market for Claude accounts in Indonesia to assess the scale of subsidy arbitrage and enforcement risks.
  • Compare Claude Fable 5.2's 'hallucination reduction' claims against independent evaluations of GPT-6 Astra and Gemini 4 Carbon.
  • Analyze the cost-benefit analysis of upgrading from the $20 to the $100 monthly plan for specific use cases like video editing.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): Ahead of every big idea, Staples Print helps Canadian businesses bring their brands to life with professional printing services that help you launch, promote, and grow. From posters and flyers to decals and menus. Staplesprint makes it easy to get your brand noticed with quality printing and support from local print experts. Shop online at
0:19
Unidentified speaker (speaker_0): staplesprint.ca or visit your nearest staples store with full service solutions to support all your business needs. Staples Business has your back. Behind every business, there's Staples Business.
0:31
Unidentified speaker (UNKNOWN): If
0:31
Unidentified speaker (speaker_1): you want to get access to this episode and my next 30 episodes, all ad free, so there'll be no ads on them. Go check out my podcast AI Chat. You can go search for that on Spotify or Apple. It's AI Chat. I'm gonna post all of these news episodes, and I'm also posting interviews like I just interviewed the CEO of Kohir. Cohere. They've raised over a billion dollars for their AI model.
0:51
Unidentified speaker (speaker_1): talking about what they're gonna be spending the money on and the direction of the AI industry, along with all of this new stuff. So if you want to go check it out with no ads for free, it is AI chat. Welcome to the podcast today. Anthropic has just launched Claude Fable 5.2 and Mythos 5.2, although we're not getting access to Mythos uh yet. I'll explain that. And they're
1:10
Unidentified speaker (speaker_1): cutting the price of Fable by 20%. There's a lot going on here. Uh Claude is Jamie's daily driver. Mine is ChatGPT, but I do have a Claude subscription. And I'll give you guys some breakdowns of some of the things I'm currently using it for. We're gonna talk about some of the ways that this new update, some of the things that it is better at, more efficient at, and why we think that will change what our current
1:30
Unidentified speaker (speaker_1): workflows are. So we'll go into some of those examples of use cases. I Uh, we'll will go into some of the examples of how we think this is changing, but this is really exciting. Before we get into these updates, with the new model from Anthropic, I wanted to mention if you haven't already joined our school community, every week we have a bonus video. Today we had a video all about how to build
1:50
Unidentified speaker (speaker_1): email lists, how to send cold outreach en masse to businesses, and how to essentially create emails that do not go to spam. It's a huge problem if you have an app, if you have a SaaS, We talk about how to make amazing, beautiful, welcome emails, intro emails, some of the tools we use and
2:10
Unidentified speaker (speaker_1): making sure none of that goes to spam. Um, as well as if you uh I talk a little bit about how I am sending 750 emails a day. I'm sure to some people, maybe that's small, but uh for one of my projects, I can get that outreach done, two cold emails, and I'm seeing some pretty incredible results. So kind of talking everything. email, and that's today's episode. But we have over a hundred episodes
2:30
Unidentified speaker (speaker_1): covering all of these different topics. You basically, in our school community, uh, can choose a path, whether being coming a consultant, creating a software, I'm in a dozen other side hustles. Our goal is for you to make ten thousand dollars extra a month from side hustles from the strategies you learn in the school community. It's 19 bucks a month. We'd love for you to join 1300 members. Link is in the description. Okay, let's get
2:50
Unidentified speaker (speaker_1): into it. what's going on with Anthropic. Jamie, so Anthropic is your daily driver. Uh, from what you're seeing here, I mean, we're seeing that this uh Fable 5.2 is going to be about 30% faster. Is this something that's gonna make an impact in some of your workflows or what you're working on?
3:07
Unidentified speaker (speaker_2): Yeah, so um I have tested out Fable 5.2 five point two just very briefly, and it does seem I
3:14
Unidentified speaker (UNKNOWN): think it
3:14
Unidentified speaker (speaker_2): seems fast. faster than Fable five point one. Uh but I I actually have not been using Fable when I use uh Cloud because they have been charging you based on usage rather than
3:26
Unidentified speaker (UNKNOWN): um you
3:27
Unidentified speaker (speaker_2): know in the the I'm just on the twenty dollar a month subscription so I don't really want to pay additional for Fable, especially when Opus five 5.5 point five has been so good. So
3:38
Unidentified speaker (UNKNOWN): um
3:39
Unidentified speaker (speaker_2): but Fable 5.2, I think for the next few days at least is
3:44
Unidentified speaker (UNKNOWN): um
3:44
Unidentified speaker (speaker_2): included in your pro subscription. So I can up to half of my usage can be 5.2. So I've been playing around a little bit with it. Um and it does seem faster. Um but I'm not gonna To like make any automations with it yet because uh in the past when there's been a test of a new model and then I
4:04
Unidentified speaker (speaker_2): lose access, I have to like redo my automation. So I haven't done anything like agentic with it. Um, but I I it does seem faster. Um, apparently, one of the claims is that if it doesn't know the answer,
4:17
Unidentified speaker (UNKNOWN): um,
4:18
Unidentified speaker (speaker_2): it will say that it doesn't know it instead of hallucinating and just making something up. So I think that's pretty cool, but I also don't know in what scenario it wouldn't know the answer to something. But I don't know. Jade, do you think this is gonna draw more users or what do you think? Like they're
4:34
Unidentified speaker (UNKNOWN): I
4:34
Unidentified speaker (speaker_2): don't know, everything just seems so similar now. Like, like is this what is this gonna change for people really at the end of the day? Oh,
4:41
Unidentified speaker (speaker_1): yeah. Well, the speed I think is pretty good. I mean, if you can get thirty 30% percent faster results, I currently have some workflows that have been running for four days straight. That's specifically on ChatGP, but I've done similar things on Claude. On ChatGPT, you can say like set a goal and then it basically sets like a long horizon goal. I think the one that I
5:01
Unidentified speaker (speaker_1): have right now that's been running for four days
5:04
Unidentified speaker (speaker_1): is it I had like 30,000 uh study guides. So it's like every verse in the Bible, I had like a study guide explaining what every verse of the Bible was about in case someone was like searching. searched like Jeremiah 7, 2, what's it about? Or like, what does it mean? And then there's like a breakdown. I used Claude to help me write all that. Well, I found out that all of those breakdowns seemed
5:24
Unidentified speaker (speaker_1): very AI-generated feeling. And so I I had like a whole list of things. I was like, remove anytime it says these like AI sounding words, these AI phrases. I kind of like created the content, but then I wanted to go audit it all. Well, running through 30,000 of those has been taking. days and days. It just runs on my computer all day long. Um, and so I think
5:44
Unidentified speaker (speaker_1): if I could cut that from, you know, 10 days down to seven days, I mean it's a meaningful difference. And it uses less tokens when it does stuff faster. So I think the speed, while it's thirty percent today, um, if you look at, you know, the next updates for the next three years of this AI model, we're gonna be these things are gonna get way faster, like it's gonna do way more for way faster. So I think the the direction
6:04
Unidentified speaker (speaker_1): of this is good. And then the other thing that I really liked was the twenty percent price cut on that. Um again, if you look over, I'm sharing my screen right now with my Claude, and I have reached a hundred percent of my weekly limit. I only have the twenty dollar Claude because I only had a couple small things, and then all of a sudden I was like, Oh man, Claude's actually really good at XYZ. So I started trying more and
6:24
Unidentified speaker (speaker_1): more things, and then of course it just burned my your twenty dollar weekly limit goes pretty fast when you're when you're getting it to get real work done. Um I'm like trying to, I'm trying to budget. I had like two $200 Claude subscriptions and one $200 Chetch P subscription. My wife told me $600 a month on this is too much. So was trying to budget. We'll see
6:44
Unidentified speaker (speaker_1): how we'll see how long we last right now. Um so anyways, yeah, I think cutting the cutting the price by twenty 20% percent is gonna make a big difference. Like because essentially what that means is you get 20% more usage out of it. So your usage goes 20% further, it's happening 30% faster. I think like while sometimes people say these are like small changes, I think they will be meaningful if
7:04
Unidentified speaker (speaker_1): we kind of compound those over a long period of time and as we do a lot of a lot of new updates in the next year or two.
7:10
Unidentified speaker (speaker_2): Yeah and I mean I think you know people are definitely using Cloud Claude for more advanced things than I am. So I could see, especially in the realm of like if you're starting to do robotics or or whatever, you know, like things like that are gonna take a lot more compute than what I'm doing. I'm gonna be uh my next kind of hustle or experiment I'm gonna be doing is actually
7:29
Unidentified speaker (speaker_2): using Claude to edit some YouTube videos, some long form. So I'm hope I'm curious to see what the usage will be like on that. Um if I'll have to bump up I bet I'll have to bump up to the at least hundred dollar a month plan.
7:41
Unidentified speaker (speaker_1): You're in twenty dollars a month?
7:42
Unidentified speaker (speaker_2): I'm on the twenty right now. Oh yeah, because I have been using
7:45
Unidentified speaker (speaker_1): stack.
7:47
Unidentified speaker (speaker_2): Yeah, I know. I'm cheap. I haven't been using AI that much recently though, but um with the uh this new thing I'm working on, I think uh I'll definitely have to bump up, but it'll save me you know hours and hours of time if it works how it's supposed to. I've seen pretty good demos online if people of ask people editing with Claude. So if you can do that for me, that's gonna be a game changer.
8:07
Unidentified speaker (speaker_1): Um, so I don't know if you saw this, but there was a recent news article that came out about a Chinese entrepreneur. I mean, just speaking about what what's being used, what's Claude's getting used for these days, there's a there's a Chinese entrepreneur, and he has reportedly been traveling um 1,500 miles to go to Indonesia.
8:27
Unidentified speaker (speaker_1): from China, and he's paying locals over in Indonesia to verify their to verify with their identity Claude accounts for him that he could go bring back to China because it's banned in China. So, anyways, it's it's pretty crazy when you have these subsidized plans, specifically the hundred like dollar, two hundred dollar a month tier
8:47
Unidentified speaker (speaker_1): on on Claude, they subsidize them, they give you so many tokens that it is it's worth some serious money, and people will literally I mean, Claude is so good. ChatGP's, in my opinion, almost just as good. Maybe this new update from Claude is now bumping it a little bit ahead. Um, but specifically the area that people are finding this doing the the best in is in
9:07
Unidentified speaker (speaker_1): all of the and all of the uh software development, which is kind of what Claude's been famous for. But they have a couple new benchmarks that they came out with um on the Frontier Code 2.0 benchmark, Fable 5.2 got a 49% on it. Opus 5.5 was the previous record holder on that, which was uh 45.5%.
9:27
Unidentified speaker (speaker_1): So, you know, I mean that's that's bumping it up. It's definitely getting better. If you compare it to models like GPT-6 Astra that got 42%, Gemini 4 Carbon got 41%. So, you know, anthropic really is leading the way with a lot of this coding stuff. You know, they're almost 50% here compared to like 42, 41% for OpenAI and and uh
9:47
Unidentified speaker (speaker_1): Google. So I will give them some credit there. Mythos 5.2, they're also going around saying how great their um benchmarks are there. I think it's way less important because um news newsflash, flash, no one gets access to Mythos 5.2 unless you're like a vetted cybersecurity. firm. So we're not all getting access to it, but inevitably, undoubtedly, Mythos
10:07
Unidentified speaker (speaker_1): 5.2 is way better at coding than than Fable 5.2. It got a 89% on the CDE rediscovery. Um, and it found 14% more known vulnerable vulnerabilities than Mythos 5.1. One. So, I mean, it's it's like 15% better at finding cybersecurity vulnerabilities than the in previous version.
10:27
Unidentified speaker (speaker_1): News Flash. None of us even got the previous version because it was too good. So it feels like the cybersecurity firms, the big tech companies, they're all getting like the juiciest, best models. And I mean, sure, I understand they got to go patch all their security before we have a have Armageddon, the end of the world with Terminator hacking everyone's breezers freezers and stuff. But um
10:47
Unidentified speaker (speaker_1): yeah, it is impressive. These models are getting better, faster, cheaper. And if peels like Anthropic is once again at the uh at the cutting edge, we'll see if OpenAI throws uh throws another towel in the ring to uh compete with them here.
11:01
Unidentified speaker (speaker_2): Thanks for listening, guys. If you got any value out of this episode, we'd really love a rating, a review wherever you are listening. Those help us out a lot and reach more people.
11:09
Unidentified speaker (UNKNOWN): And
11:10
Unidentified speaker (speaker_2): again, check out the AI Hustle school community if you want to learn how to take some of these tools and actually make some money with them and grow your business. I would love to have you be a member. And so thanks for listening. We'll see you next time.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from AI Investing: for the AI Investor

AI Investing: for the AI Investor

OpenAI Launches ChatGPT Wallet and More AI News

Episode discussion: OpenAI launches ChatGPT Wallet, SpaceX announces $74B AI campuses, and NVIDIA CEO resigns amid spy allegations

Speakers’ claims: Speakers discuss OpenAI's launch of the ChatGPT Wallet, a feature allowing US Pro subscribers to make purchases via AI agents using tokenized cards without transaction fees. The episode covers Google's pre-order launch of Gemini smart glasses priced at $400 and Cerebrus's record Q3 revenue of $230 million. A significant portion focuses on OpenAI backing a bill to tighten Chinese AI chip export controls following NVIDIA CEO Jensen Huang's resignation amid allegations involving a Chinese spy. SpaceX announced plans for over $74 billion in AI supercomputing campuses in Texas and Arizona. Finally, Mirandil raised $1 billion for its coding model, demonstrating efficient model development.

OpenAINVIDIASpaceX
18 min4 assets10 claims to verify
AI Investing: for the AI Investor

Claude Fable 5.2 Is Here — Plus NVIDIA's Jensen Huang Investigation and Amazon's Home Robot

Episode discussion: Anthropic cuts Claude pricing and boosts speed; NVIDIA faces CEO probe; Amazon builds robotics unit

Speakers’ claims: The episode covers Anthropic's launch of Claude Fable 5.2, which offers a 20% price cut and 30% speed increase, scoring 49.1 on Frontier Code 2.0. NVIDIA's board launched an investigation into CEO Jensen Huang over allegations involving a Singapore-based executive, causing an 8% stock drop. Amazon is forming a personal robotics group around Fauna Robotics' Sprout robot to compete with Tesla. Apple faces internal pressure to accelerate AI features for iOS 27, while Samsung tests a $4.99/month AI subscription. DeepSeek reports rapid user growth for its Harness Agent app, and Elon Musk addresses Grok token limits.

AnthropicNVIDIAAmazon Robotics
18 min5 assets10 claims to verify
All insights