👋 Hello hello,
Anthropic shipped its most powerful model ever, Fable 5. Three days later, the US government export-controlled it, blocking access for every foreign national on the planet. That forced Anthropic to pull it for all of us.
So we've got the strangest flex in tech right now: the most-wanted AI model in the world is one almost nobody can legally touch. Naturally, the internet spent the weekend trying to rebuild it from spare parts.
Let's get into it.
🔥🔥🔥 Three Curated AI Updates
Anthropic released Fable 5 this month, the first model in its new Mythos-class tier that sits a notch above Opus 4.8 in raw capability. It lasted barely a week. The Commerce Department sent Anthropic a letter placing Fable 5 and its bigger sibling Mythos 5 under export controls, blocking any foreign national inside or outside the US, including Anthropic's own non-citizen staff.
Because the company couldn't selectively lock out foreign nationals, it switched both models off for everyone. Every other Claude model, Opus 4.8 included, still works fine.
The reported trigger was a "jailbreak" that used Fable's cybersecurity skills to find software flaws. Anthropic says the evidence was one verbal demo that surfaced minor, already-known bugs other public models can find too, and it's contesting the order. Worth watching closely, since this is the first time Washington has export-controlled an AI model itself instead of the chips behind it.
Quick context: Perplexity's Deep Research is an agent that runs dozens of searches, weighs the sources, and writes up an answer the way a research analyst would. Computer is its workspace that turns answers into real deliverables like decks, dashboards, and sites.
Until now those lived in separate threads. You'd research in one place, then rebuild the output somewhere else. Perplexity merged them, so a single prompt can go from hard question to finished report, pulling in your connected files and premium sources like Statista and PitchBook along the way.
Under the hood it uses a "Search as Code" approach, writing a custom search program that fires thousands of lookups in parallel. For anyone who does desk research for a living, that's a real time-saver.
This one's pure fun. Google quietly switched on a flight simulator inside Google Earth on the web, free for everyone, no download. Open the Tools menu, pick a starting point, and fly over real 3D terrain with your keyboard and mouse.
It's not a serious sim and the controls take a minute to click, but buzzing your own neighborhood from a cockpit is a genuinely good time. Fun fact: this started as a hidden Easter egg back in 2007, unlocked with a secret keyboard combo. Now it's a menu click for the whole planet.
Where would you fly first?
🔥🔥 Two Fable Replacements, Maybe?
Security researcher Jamieson O'Reilly ran a clever test: how much of Fable's magic lived in the model itself versus the hidden system prompt that shapes its personality? He layered the leaked Fable prompt on top of Opus 4.8 and compared it against plain Opus 4.8.
The useful takeaway for anyone who prompts AI daily: a system prompt can change how a model talks and carries itself, but it can't hand you capabilities the base model doesn't have. You essentially get Opus 4.8 with a Fable accent.
Fusion takes a different route. Instead of copying Fable, it sends your prompt to a panel of several models at once, has a "judge" model compare their answers for agreement and gaps, then synthesizes one final response. OpenRouter says this matched Fable 5 on a 100-task deep-research benchmark at roughly half the cost. It's best for hard research questions where blending models beats betting on one.
Two honest caveats: the exact panel that tied Fable actually used Fable, which is now offline, and the benchmark tested zero coding tasks. Still a smart patch for the Fable-shaped hole. Call it with the slug openrouter/fusion.
🔥 Things You Didn't Know You Could Do With AI
🎬 Opus 4.8 vs. Fable 5 for video editing
More people are already using AI as a video editor, and not just for captions and trims. We're talking restructuring a clip for better pacing, with more dynamic motion and interactive touches.
Before Fable 5 got pulled, we ran a little test. We took one of Kushank's videos and had both Opus 4.8 and Fable 5 edit the same raw clip, then compared them side by side. The results were closer than you'd expect. Both delivered a clean, watchable cut. Fable 5 had the edge, with a more dynamic feel and more interactive elements, the kind of polish you'd usually add by hand.
The practical part: Fable's offline right now, but Opus 4.8 still does a genuinely solid job and it's available today. If you've got raw footage sitting around, it's worth seeing how far an AI edit gets you before you ever open a timeline.
Don't forget to rate today's post
Until next time,
Team @PracticalyAI
