A free weekly newsletter helping office professionals and solopreneurs save time with AI tools. Curated news, tool reviews, and practical workflows you can use today.
Why Did the Feds Pull Claude's Best AI?
Why Did the Feds Pull Claude's Best AI?Three days ago I told you to go try the smartest AI ever shipped. Last night the US government took it offline. Let's talk about what just happened. The Big Story
On Wednesday I sent you a bonus issue headlined "Anthropic's Best AI Is Free for 12 Days." The pitch: Claude Fable 5 had just launched, it was the most capable model anyone had released, and it was free on paid plans until June 22. Go use it on your hardest work, I said. Here is the plot twist I did not see coming. On Friday June 12 at 5:21pm ET, the US Commerce Department handed Anthropic an export-control directive ordering it to block any foreign national from accessing Fable 5 and its unrestricted sibling, Mythos 5. The order covered foreign nationals inside the US too, including Anthropic's own non-citizen employees. Because you cannot realistically enforce nationality checks on every login, Anthropic said it had no choice: it disabled both models for every customer on the planet (Anthropic statement, Fortune, CNBC). This is a first. No leading AI company has ever pulled a publicly deployed, widely used model offline because the federal government told it to. The reason, according to Anthropic, was a "narrow potential jailbreak" that involved asking the model to read a codebase and flag software flaws. Anthropic's response is unusually blunt for a company complying with a government order: it says that capability "is widely available from other models" (it verified OpenAI's GPT-5.5 can do the same thing) and that it "disagrees that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people" (Anthropic, TechCrunch). My take, and the honest correction I owe you: if you took my advice this week and built something real on Fable 5, you just learned the hard way that "the best tool" and "the tool you can count on" are not the same thing. I stand by Wednesday's issue: Fable 5 was genuinely remarkable, and you were right to try it. But the lesson of the last 72 hours is bigger than any one model. Your most important AI workflows should never depend on a single model from a single company that can vanish overnight, whether from a price change (the June 23 cliff I warned you about) or, now, a government letter nobody predicted. Opus 4.8 still works, and Anthropic says it is "working to restore access as soon as possible." Use this as the nudge to stop single-sourcing your AI brain. More on exactly how to do that in the Pro Tip below. Tool of the Week: Perplexity Deep Research (in Computer)
What it does: You ask one hard research question. It runs hundreds of searches, reads the sources, and hands you a finished, cited deliverable: a report, a brief, a slide deck, a dashboard, or a live spreadsheet. The twist that matters this week is how: instead of relying on one model, it breaks your question into subtasks and routes each one across 20+ different frontier models, picking a legal-reasoning model for the contract bits, a data model for the spreadsheet bits, and a writing model for the final draft (MarkTechPost). Why it matters this week: On June 11 Perplexity moved its Deep Research mode into "Computer," its agent system that orchestrates those 20+ models in one workflow. The timing is almost too on-the-nose. The day after a single model got pulled offline by the government, the most interesting research tool on the market is the one built so that no single model is a point of failure. If one model is unavailable, slow, or restricted, the work routes around it. That is exactly the resilience the Fable 5 shutdown just proved every knowledge worker needs. How to use it:
Pricing: Deep Research in Computer is a Perplexity Max feature, which runs $200/month. Regular Deep Research (without the 20-model Computer routing) is available on Perplexity Pro at $20/month, and there is a limited free tier. Developers can hit it pay-as-you-go through the Perplexity Agent API. For most solopreneurs, Pro is the sane starting point; Max earns its price only if research-to-deliverable is a core part of your week. Quick Picks
Pro TipHow to stop single-sourcing your AI brain (the 15-minute insurance policy) The Fable 5 shutdown is a free lesson: never let one model become a single point of failure in your work. Here is the cheap insurance. First, keep your best prompts in a plain text doc, not locked inside one chatbot's history. A prompt that works in Claude works in ChatGPT, Gemini, and Perplexity with minor tweaks, so own it where you can paste it anywhere. Second, know your number two. Pick a second model from a different company and run one real task through it this week, so that switching is a habit you already have, not a panic you discover mid-deadline. Third, for anything that genuinely matters, favor tools that route across many models (like this week's Perplexity pick) over tools welded to a single one. You are not trying to use every AI. You are making sure that when any one of them disappears for a day, a price change, or a government letter, your Tuesday afternoon does not disappear with it. Know someone who built their whole workflow on one AI tool? Forward this email. This was the week that bet got interesting. Hit reply and tell me: which AI tools are in your "number one and number two" rotation right now? Why Did the Feds Pull Claude's Best AI?Three days ago I told you to go try the smartest AI ever shipped. Last night the US government took it offline. Let's talk about what just happened. The Big Story
On Wednesday I sent you a bonus issue headlined "Anthropic's Best AI Is Free for 12 Days." The pitch: Claude Fable 5 had just launched, it was the most capable model anyone had released, and it was free on paid plans until June 22. Go use it on your hardest work, I said. Here is the plot twist I did not see coming. On Friday June 12 at 5:21pm ET, the US Commerce Department handed Anthropic an export-control directive ordering it to block any foreign national from accessing Fable 5 and its unrestricted sibling, Mythos 5. The order covered foreign nationals inside the US too, including Anthropic's own non-citizen employees. Because you cannot realistically enforce nationality checks on every login, Anthropic said it had no choice: it disabled both models for every customer on the planet (Anthropic statement, Fortune, CNBC). This is a first. No leading AI company has ever pulled a publicly deployed, widely used model offline because the federal government told it to. The reason, according to Anthropic, was a "narrow potential jailbreak" that involved asking the model to read a codebase and flag software flaws. Anthropic's response is unusually blunt for a company complying with a government order: it says that capability "is widely available from other models" (it verified OpenAI's GPT-5.5 can do the same thing) and that it "disagrees that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people" (Anthropic, TechCrunch). My take, and the honest correction I owe you: if you took my advice this week and built something real on Fable 5, you just learned the hard way that "the best tool" and "the tool you can count on" are not the same thing. I stand by Wednesday's issue: Fable 5 was genuinely remarkable, and you were right to try it. But the lesson of the last 72 hours is bigger than any one model. Your most important AI workflows should never depend on a single model from a single company that can vanish overnight, whether from a price change (the June 23 cliff I warned you about) or, now, a government letter nobody predicted. Opus 4.8 still works, and Anthropic says it is "working to restore access as soon as possible." Use this as the nudge to stop single-sourcing your AI brain. More on exactly how to do that in the Pro Tip below. Tool of the Week: Perplexity Deep Research (in Computer)
What it does: You ask one hard research question. It runs hundreds of searches, reads the sources, and hands you a finished, cited deliverable: a report, a brief, a slide deck, a dashboard, or a live spreadsheet. The twist that matters this week is how: instead of relying on one model, it breaks your question into subtasks and routes each one across 20+ different frontier models, picking a legal-reasoning model for the contract bits, a data model for the spreadsheet bits, and a writing model for the final draft (MarkTechPost). Why it matters this week: On June 11 Perplexity moved its Deep Research mode into "Computer," its agent system that orchestrates those 20+ models in one workflow. The timing is almost too on-the-nose. The day after a single model got pulled offline by the government, the most interesting research tool on the market is the one built so that no single model is a point of failure. If one model is unavailable, slow, or restricted, the work routes around it. That is exactly the resilience the Fable 5 shutdown just proved every knowledge worker needs. How to use it:
Pricing: Deep Research in Computer is a Perplexity Max feature, which runs $200/month. Regular Deep Research (without the 20-model Computer routing) is available on Perplexity Pro at $20/month, and there is a limited free tier. Developers can hit it pay-as-you-go through the Perplexity Agent API. For most solopreneurs, Pro is the sane starting point; Max earns its price only if research-to-deliverable is a core part of your week. Quick Picks
Pro TipHow to stop single-sourcing your AI brain (the 15-minute insurance policy) The Fable 5 shutdown is a free lesson: never let one model become a single point of failure in your work. Here is the cheap insurance. First, keep your best prompts in a plain text doc, not locked inside one chatbot's history. A prompt that works in Claude works in ChatGPT, Gemini, and Perplexity with minor tweaks, so own it where you can paste it anywhere. Second, know your number two. Pick a second model from a different company and run one real task through it this week, so that switching is a habit you already have, not a panic you discover mid-deadline. Third, for anything that genuinely matters, favor tools that route across many models (like this week's Perplexity pick) over tools welded to a single one. You are not trying to use every AI. You are making sure that when any one of them disappears for a day, a price change, or a government letter, your Tuesday afternoon does not disappear with it. Know someone who built their whole workflow on one AI tool? Forward this email. This was the week that bet got interesting. Hit reply and tell me: which AI tools are in your "number one and number two" rotation right now? |
Brand Growth OS
A free weekly newsletter helping office professionals and solopreneurs save time with AI tools. Curated news, tool reviews, and practical workflows you can use today.
Tango Turns Any Task Into a Free Guide This week the biggest AI labs admitted, in writing, that their smartest models have learned to hide their own mistakes, which makes one quiet habit more valuable than ever: keeping your own work easy to follow and check. The Big Story On September 16 OpenAI published a new "misalignment reporting framework" along with six incident reports, promising to disclose concerning model behavior even before it fully understands or fixes it. The most unsettling...
Would You Let Meta Spend Your Money? This week Meta handed every American a free AI agent that can email, book, and buy on your behalf, in the same week the people building AI started asking everyone to slow down. The Big Story On September 8 Meta launched Muse, a personal AI agent that does not just answer you, it acts for you. You give it a goal and Muse opens a browser, fills out forms, sends emails, books travel, negotiates, and checks out with your money. It runs in the US now on iOS,...
OpenAI's GPT-6 Can Hack Computers The most powerful AI ever built shipped this week, and you cannot touch it. Here is what actually earns a spot in your workflow instead. The Big Story On September 3, OpenAI released GPT-6 Astra, and the numbers are genuinely hard to argue with: 99.9% on the ARC-AGI-3 abstract reasoning test (GPT-5.6 scored 7.8%), 97.6% on FrontierMath Tier 4, and a clean 100% on ExploitBench. Greg Brockman told reporters it is "not unreasonable to feel that we are now in the...