There’s been a lot of back and forth about the unveiling of Anthropic’s Fable model as a concession or alternative to the walls garden for Mythos (about 12 tech firms are the only authorized users of the Mythos model) – but as Ethen Mollick points out in his newest post on his blog, One Useful Thing, on June 9, Fable essentially is Mythos – with some guardrails built in.

To recap what’s been going on, Anthropic put out its new Mythos model early in April, but restricted it to Cisco and a handful of other U.S. companies, because of extreme concerns about cybersecurity. U.S. government officials had been meeting with the banks and talking about how Mythos is powerful enough to undermine important systems, so stakeholders created something called Project Glasswing to protect the world from the model’s awesome power.

So just a short while ago, Anthropic came out with Fable, or basically, Mythos defanged. The mechanism of Fable’s safety protocol is interesting, too. Apparently, at the first whiff of a cybersecurity issue, the model hands a request off to one of its lesser peers (Mollick names Claude 4.8 Opus, but I’m getting ahead of myself) so that the model doesn’t turn into a do-it-yourself platform for hackers, and start to infiltrate the best security wrapping sensitive networks.

Ethan Mollick is one of the most interesting voices on AI, always on the cusp of new model research. I follow his blog keenly, as he documents all of the iterative progress we see with LLMs. He also has MIT connections, and has posted a prodigious amount of material on YouTube.

Anyway, Mollick has been playing around with Fable, not in a cybersecurity context, but otherwise. He explains:

“Much of the discussion of Mythos has centered on its impact on software security,” Mollick writes, “but I tested it on everything except that (the guardrails around Fable essentially prevent it from being used for cybersecurity at all). My conclusion is that it represents a very real leap over every model I have used before, and, maybe more important, suggests our relationship with AI is changing in drastic ways.”

One thing that comes across, to me, in reading the rest of the blog post, is that Mollick is inviting us to try out his creations, to see, in a sense, what he saw as he tested Fable’s prowess.

“The results were impressive,” he writes, of something called an isochrone map that shows travel distances from London, a project which, he explains, is impressively complex. “I pushed a few more times in directions that interested me (including asking for other visualization approaches, etc.). I would recommend spending a couple minutes clicking around the results, and you can read its methods and sources at the bottom of the graph.”

Mollick also used Fable to put together a few games, and some kind of alliterative poetry about haircuts, and he showcases these results, to provide examples of what Fable, as today’s cutting-edge model, is able to do.

Another very interesting point that Mollick brings up is that although Claude can’t technically produce images using standard methods like Stable Diffusion, Fable somehow comes up with passable results through code.

“I … had (Fable) create a bunch of games you can try,” Mollick writes. “All of these are one initial prompt in Claude Code, where Fable had to take my vague prompts and generate something workable, followed by a couple of additional prompts with minor encouragement or feedback. What makes these especially impressive is that Claude cannot generate images, so every piece of art or 3D object was made with math alone, not using any external assets.”

Obviously, creating images with code is the long way around. Others have figured out other ways to get around the image problem with Claude, for instance, using APIs or MCP protocol.

Again, Mollick marvels about the model’s capabilities, though still acknowledging that “jagged frontier” that remains part of the equation.

“The output is impressive,” he writes. “But, especially as I turned to more serious projects, I often felt using the tool was somewhere between delightful and unnerving. Delightful, because I just asked for something, and it happened. And also unnerving, because I just asked for something, and it happened.”

Here’s another way that Mollick expressed that same wonder, in an endorsement of Fable’s powers that seems pretty robust.

“The deeper strangeness is how little I had to do, and how little I could see while it was being done,” he writes. “Last year, I called this working with a wizard: you chant the spell and something happens. With Fable, the spell has gotten powerful enough that I am no longer sure I am the wizard. I am closer to a patron. I describe what I want, I pay for it, and I judge the result. The conjuring happens somewhere I cannot watch, in hundreds of small choices I never get a vote on. The work has shifted from process to outcome. I no longer steer; I commission.”

It sounds a lot like vibecoding…

“A patron commissions a single artist. Fable is closer to a whole studio, where I am the client who signs off on the final work without ever setting foot on the floor.”

I think a lot of this stands on its own, and invites the rest of us to really think about what working with these models will be like in 2027. Which is approaching. I spent the first half of this year talking about how 2025’s realities led into 2026: now we see the next phase emerging. Safety first.