22:24Riley Brown · Tutorial
Claude Fable Will Change EVERYTHING (Here's Why)
A 22-minute breakdown of Anthropic's Mythos-class Fable 5 model — live demos, a structural argument about building blocks, and a pricing window closing June 22.
June 12thThe argument in one line.
A US government export-control directive — citing a reported jailbreak with no specific details — pulled the most capable public AI model ever released within 76 hours, establishing that governments can revoke frontier model access on national security grounds with virtually no notice.
Read if. Skip if.
READ IF YOU ARE…
- You use Claude and woke up to find Fable 5 unavailable and want a clear account of exactly what happened.
- You are building on the Claude API and need to know which models are affected and what Anthropic's official stance is.
- You are tracking AI regulation and want to understand the first documented government pull of a commercial frontier model.
- You are curious about the real cost economics of running Mythos-class models at API pricing.
SKIP IF…
- You want a technical analysis of the jailbreak — neither the video nor Anthropic's statement provides it.
- You want a legal breakdown of the statutory authority the government used.
The full version, fast.
Seventy-six hours after releasing Claude Fable 5 — the public, safety-guardrailed version of Mythos — Anthropic was served a US government export-control directive suspending all global access, including for US customers. The directive cited a reported jailbreak by an unnamed company but gave no specific details. Anthropic complied under protest, calling it a misunderstanding and stating it violates their own published standard for transparent, technically-grounded government oversight. Only Fable 5 and Mythos 5 are affected; Haiku, Sonnet, and Opus 4.8 remain available. Full API pricing for Fable 5 ran an estimated 600 dollars per hour — over 1.2 million dollars per year at a 40-hour workweek.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:12
01 · Breaking news and the unavailable message
Host announces the ban, shows the in-app error message, notes global scope including US users and Anthropic employees.
01:13 – 02:07
02 · What is Fable / Mythos?
Star-rating comparison: Mythos = 5 stars, Fable 5 = 4 stars (safeguards applied). Explains Anthropic nerfed Mythos for public release.
02:08 – 02:55
03 · The jailbreak claim
An unnamed company reportedly bypassed Fable 5 safeguards; government cited this as national security justification with no specifics.
02:56 – 03:57
04 · Anthropic's official statement
Directive received at 5:21 ET with no specific details. Anthropic complies under protest, calls it a misunderstanding, plans to restore access.
03:58 – 05:01
05 · Community reaction
Developers cut off mid-build. Theo's viral eulogy tweet. Todd Saunders demo of Mythos building a full product during a live customer call.
05:02 – 05:44
06 · What Mythos could actually do
Lovable mobile clone in 2 prompts, humanoid robot design, F-Zero racing game, spatial planning app. Positioned as the most fun model ever.
05:45 – 06:16
07 · Was this Anthropic's plan? The IPO theory
Some X users claim Anthropic used the government as a scapegoat to sunset an unsustainably expensive model before an IPO. Presented without endorsement.
06:17 – 07:49
08 · The true cost of running Fable 5
600 dollars per hour on API pricing, 1.248 million per year at full-time use. Subscription cutoff already set for June 22. Host's own Lovable clone cost 200-300 dollars in 9-10 prompts.
07:50 – 08:14
09 · Which models are affected
Only Fable 5 and Mythos 5 suspended. Haiku, Sonnet, Opus 4.8 fully available. Illustrated on whiteboard diagram.
08:15 – 09:04
10 · The bigger stakes
Comparison to Sam Altman firing. Dystopia framing: what if only governments hold frontier AI access going forward?
09:05 – 10:58
11 · Sponsor: Chorus.com
Live demo of Chorus — iMessage-based agent running Claude Code in the cloud. Builds a Theo Brown research landing page via voice message in real time.
Lines worth screenshotting.
- A US government directive pulled the most capable public AI model in history from all global customers — including US users — with no specific justification and 76 hours after launch.
- Fable 5 was Mythos with safety guardrails; the jailbreak concern was that those guardrails had been bypassed, effectively restoring full Mythos capability in adversarial hands.
- Anthropic explicitly disagrees with the government's reasoning while complying — legal pressure can override a company's own safety posture on its own product.
- Running Fable 5 at full API pricing costs roughly 600 dollars per hour — 1.2 million dollars per year at a 40-hour workweek — making it financially inaccessible to nearly every commercial user.
- Anthropic's own published standard for government AI oversight requires transparency, fairness, and technical grounding; the directive it was forced to comply with meets none of those criteria.
- Haiku, Sonnet, and Opus 4.8 are completely unaffected — the ban is model-specific, not a company-wide suspension.
- Claude subscribers who paid 100-200 dollars per month were already slated to lose Fable 5 access on June 22; the government directive simply accelerated that cutoff.
- The precedent set here — government revocation of a commercial AI model on unspecified national security grounds — applies equally to any frontier model deployment.
- Community reaction ranged from developers mid-build losing access without warning to executives comparing the disruption to Sam Altman's 2023 firing.
- The scenario where only governments and approved companies access frontier AI models is no longer hypothetical — it happened briefly to every Fable 5 user simultaneously.
The first government model-pull sets a precedent every AI user should understand.
WHAT TO LEARN
When a government can suspend the most capable public AI model with 76 hours notice and no specific justification, every developer and user bears that same exposure.
- Anthropic complied with a directive it explicitly disagrees with — legal pressure can override a company's own safety posture on its own product.
- The directive applied globally including US customers, meaning domestic product status provides no protection against export-control-style orders.
- Fable 5 was the safety-guardrailed public version of Mythos; the jailbreak concern was that those guardrails were bypassed, reverting capability to Mythos-level in adversarial hands.
- Full API pricing for Fable 5 ran roughly 600 dollars per hour — 1.2 million dollars per year at a 40-hour workweek — meaning the model was financially inaccessible to most commercial users even before the ban.
- Anthropic's own published standard for government AI oversight (transparent, fair, technically grounded) was violated by the directive it was forced to comply with — a tension the company flagged publicly.
- Haiku, Sonnet, and Opus 4.8 are unaffected — the ban is model-specific, not company-wide, which limits immediate damage but does not reduce the precedent.
- The scenario where only governments and approved companies access frontier AI is no longer theoretical — it happened simultaneously to every Fable 5 user worldwide.
Terms worth knowing.
- Claude Mythos
- Anthropic's most powerful internal model, described as the most capable AI system in the world; restricted to enterprise partners and government entities before any public version.
- Claude Fable 5
- The public-facing, safety-guardrailed version of Mythos released to all Claude subscribers; slightly below Mythos in raw capability due to built-in safety classifiers.
- Export control directive
- A US government order under national security authority requiring a company to restrict access to a technology — here applied to an AI model at commercial scale.
- Jailbreak
- A technique that bypasses an AI model's built-in safety constraints, allowing it to respond to requests it would normally refuse; cited by the government as justification for the directive.
- Safety classifier
- A secondary model or rule set layered over a base AI that filters or blocks unsafe outputs; Fable 5 had classifiers Mythos lacked, and the reported jailbreak allegedly nullified them.
Things they pointed at.
Lines you could clip.
00:00
“Anthropic just banned all access to Claude Fable 5 — the safe version of Mythos — 76 hours and 20 minutes after releasing it to the public.”
03:06
“The letter did not provide specific details of its national security concern.”
07:30
“At a forty-hour work week, that's 1,200,000 dollars per year just on Mythos tokens.”
08:28
“If we were to lose access as common citizens of the world — to the best AI models — and only governments and certain companies had access, that sounds like a pretty terrible world to live in.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
analogystory
00:00Ladies and gentlemen, we have some breaking news. Anthropic just banned all access to Claude Fable five, which was the safe version of mythos that they released to the public seventy six hours and twenty minutes ago. The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable five and Mythos five by any foreign national whether inside or outside The United States including foreign national anthropic employees.
00:30The net effect of this order is that we must abruptly disable Fable five and Mythos five for all customers, not just foreign customers in The US as well to ensure compliance. Almost everyone I've talked to, even the people deep in the industry are very confused by this situation.
00:48Other people like Aaron Levy are saying that this is a turning point for AI regulation. The government is starting to deem some models too powerful for certain uses which creates a precedent for a range of possible controls in the future. And no matter where you try and use Claude, whether that is the Claude desktop app or Claude co work or Claude code, you're gonna get hit with one of these messages when you try to use Fable five high.
01:13This model isn't available right now. You can switch to another model to continue using Claude. Alright.
01:19I'm gonna do my best to answer all of the questions that you may have about Claude banning all access to Claude Fable five. And so the first question you may have is what is Claude Fable or what is Claude Fable five? And for those of you who don't know, Mythos has been a model that Anthropic has been teasing for a long time, and they gave early access to Mythos to a lot of big companies and the government.
01:42And so Claude Mythos is like the most powerful model in the world. It is absolutely insane.
01:48Five stars in terms of power. Then they released a public version of Mythos which is Claude Fable five and I put it here as four star power because they took away some of the power manually by imposing some safeguards.
02:03So Anthropic basically nerfed this model by putting up some safeguards. So if you ever try to do some frontier LLM research with Claude Fable five, it'll automatically nerf you and it may even reroute you to a different model.
02:19And so the reason this model is no longer inside Claude for us to use is that the US government is basically saying that some company and they're not even naming which company were able to jailbreak Fable five and get it to do whatever they want. So they're basically removing this safeguard and in theory they can get the model to do whatever they want and we don't know what we don't have all the details.
02:42We just know that this is what the government said and Anthropic has been kind of forced to comply. And so Anthropic was indeed basically forced to do this.
02:53They are complying with a legal directive but they disagree with the government's reasoning. They don't believe they should have had to cut off access to Fable five.
03:02And the weird part is according to the official statement by Anthropic, they said that we received the directive from the government today at 05:21 eastern time.
03:12The letter did not provide specific details of its national security concern. Our understanding is that the government believes it has become aware of a method of bypassing or jailbreaking fable five.
03:25And Anthropic has recently said that the government should have the ability to block unsafe deployments as part of statutory process that is transparent, fair, clear, and grounded in technical facts.
03:38This action does not adhere to those principles, so they are actively against this government decision. We believe that this is a misunderstanding, and we are working to restore access as soon as possible.
03:48So, hopefully, in the next twelve hours or twenty four hours or thirty six hours, we do have access. But as of right now, we simply do not have access. And it's just a crazy time.
03:59I've talked to five different people at different companies who said that they were working on something incredibly important to their business using Fable five, spending a ton of money on tokens, and in the middle of it, it just shut off out of nowhere.
04:14And Theo said, Fable, my beloved, I will miss you so. Our three days together were magical unlike anything I've experienced before it.
04:22Some things are too good to be true. So good that the government interferes. I'm sorry we were one of those things until we meet again.
04:29And the reason this is so sad is Mythos was clearly a step up. It was better than all of the previous models that came before it. And as Todd Sanders explains here, he said that he was on a customer call today and Claude was transcribing in the background as they were telling me about the features they wish they had with their current software.
04:47Claude was building the features in real time. By the end of the call, I was able to show a fully working product with the exact workflow they mentioned fifteen minutes earlier. Autonomous looped building triggered from a customer call.
05:02And I've talked about this already, but I fully rebuilt Lovable, their mobile app with Claude Mythos. In a single prompt, Claude Mythos was building perfect Minecraft clones. It even built this app that allows people to, like, plan out their video sets nearly perfectly and it's incredible at spatial awareness on Fable five could also fully design these humanoid robots and also create these incredibly realistic racing games as Peter Yang did yesterday.
05:35And I've basically paid close attention to every single model drop and I'm telling you people had more fun with this model than any model previously. And so the most powerful, most fun, and most useful AI model ever released is now just gone. So many people are saying that this was Anthropic's plan all along.
05:55They're saying that there's no way that it was sustainable to run Mythos for everyone given the amount of computing power it required and the computing, uh, resources that they had. It just wasn't sustainable. So instead they planned this.
06:09Right? They released Fable to the public, get people excited, and then take it away.
06:14And they're using the government as a scapegoat. And this model is heavy.
06:19It's very expensive to run and it's very expensive to use. Earlier today, you tried to click on Fable five in your Claude account which means your Claude subscription, so whether you pay $200 month, a $100 month, or $200 a month, you would see that it's only available in your Clod account until, like, June 22. That no longer shows up here because it's literally just a banned model.
06:40You just get hit with this message right here. Even prior to Claude banning the model today, people were already going to lose their access on June 22 and then you would have to pay for Claude Fable five through API pricing which is way more expensive.
06:56You have to pay for all of your usage. And this guy tweeted that it is about $600 per hour to run Mythos.
07:04At least however he was running Mythos, maybe he was running two Mythos sessions in parallel but for him it was $600 an hour. And I'm telling you, if you're running Mythos seriously, it is anywhere from 300 to $600 an hour on API pricing.
07:19In order for me to create my lovable clone, it did cost around 200 to $300 and that was only like nine to 10 prompts. And so if you it is $600 per hour at a forty hour work week, that's $1,200,000 per year just on Mythos tokens if you're running it all the time for those hours.
07:42And so this really expensive model Claude Fable five which is a derivative of Claude Mythos that was meant to have some safeguards, it has been removed. That is the only model that is affected if you try to use Claude right now you can still use Haiku, you can still use Sonnet and you can use Opus 4.8 just like normal.
08:03I'm going to be posting about this on my Twitter and I'll even be posting shorts on my Instagram as I learn more. This is a huge deal. I just wanna put this video out.
08:13Hopefully, it gets restored very soon. Let me know in the comments if I missed anything. Let's have a discussion in the comments.
08:20I have not I don't remember something that was such a big deal since maybe late twenty twenty three when Sam Altman was fired from OpenAI. This feels like an equivalent thing. And if we were to lose access as as common citizens of the world, if we lost access to the best AI models and only governments and certain companies had access, that sounds like a pretty terrible world to live in.
08:44It's really sad as someone who's been deep in the AI space for three straight years since ChatGPT came out. To lose access to Frontier models sounds really sad and really depressing and I really hope we get access soon.
08:57And guys, by the way, if you wanna add Claude code and codex to your iMessages without needing to use a physical computer, you could just go to chorus.com and in one click, it'll just give you a phone number that you can text and you can immediately start using the world's best models directly in iMessage. This is mycourse.com content agent and has access to YouTube.
09:16It can scrape the transcripts from any Instagram or TikTok video. Hey, buddy. I need you to research the latest videos by Theo Brown and tell me what he's been saying about Anthropic and OpenAI.
09:28Give me a summary. I want you to make a little landing page, just like a little public link so I can just browse it on the web here. And yeah.
09:34Thank you. And see it just responded. It says, on it, digging into Theo's latest, pulling what he said about Anthropic and OpenAI, and it says a couple minutes.
09:44Love it. Research landing page building right now. Cool.
09:47And while it's loading, can show you that you can very easily just go to chorus.com and click skills. And you can see here that I have it connected to YouTube, which is my YouTube account.
09:56And then there's also just massive marketplace of really useful skills like SupaData, which allows you to pull transcripts from any of the major platforms. That's one thing we wanna do is make the best marketing skills available and allow you to message it directly through iMessage.
10:11And take a look at that. It is done. And so I'm gonna go ahead and open this up right here.
10:17And this is fully Cloud Code running in the cloud, so it can build any type of app, web app, mobile app, landing page, doesn't matter. It can create any type of app. And here we go.
10:27We have all of his videos. And we have this massive research report. This was created by Opus four point eight, and it analyzed every single one of Theo's videos where he mentioned Anthropic and Claude code.
10:40And you could try this and have this set up in about one minute. Literally one minute you could have this set up. Our team and I, we've been working on this for the past two months on chorus.com and you come and you can immediately just start texting your agent.
10:52Thank you so much for watching. I hope you guys got a ton of value out of this. I'll see you here for the next one.
10:57Peace.
The Hook
The bait, then the rug-pull.
Seventy-six hours and twenty minutes after Anthropic made the most powerful public AI model in history available to subscribers, the US government made it disappear. What followed was a cascade of mid-build shutoffs, six-figure token bills cut to zero, and an official statement that amounts to: we complied, but we think you are wrong.
Named ideas worth stealing.
01:13model
Mythos vs Fable capability tiers
- Mythos (5 stars — full power, enterprise/government only)
- Fable 5 (4 stars — public, safety classifiers applied)
Star-rating visual showing how Anthropic tiered model access: Mythos for approved partners, Fable 5 as the safety-limited public version.
Steal forAny explanation of tiered model access or product vs API capability gaps
How they asked for the click.
VERBAL ASK
09:05product
“If you wanna add Claude code and Codex to your iMessages without needing to use a physical computer, you could just go to chorus.com”
Live demo running in real-time — builds a web research page via iMessage voice prompt while talking. Strong because it shows the product working, not just describes it.
MENTIONED ON CAMERA
Visual structure at a glance.
