Ethan Mollick's early test of Anthropic's Mythos-class model shows it tackling a roughly 19-page design spec with almost no ongoing guidance, running for 9.5 hours while spinning up sub-agents, pulling real flight and rail data, writing code, and verifying its own outputs to deliver working software for statistical analysis.
I've had access to Fable for a bit. A genuine jump in capability, I could feed it a 15 page design document for a project and it would work for 9+ hours and deliver terrific results.
But working with it is weird & weirder is coming
Lots of examples: https://open.substack.com/pub/oneusefulthing/p/what-it-feels-like-to-work-with-mythos?r=i5f7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
Instead of constant back-and-forth, the model handles hundreds of micro-decisions on its own and only surfaces occasional updates, which feels novel for anyone used to steering AI step by step.
At double the per-token rate of earlier Opus models, the nine-hour runs could get expensive unless cheaper sub-agents offset much of the spend, though exact totals from the tests remain unclear.
Many users praised Fable AI completing complex projects over nine hours and the article's snake game example for its impressive results, while others dismissed the claims as uninnovative or questioned safety priorities.
Some fun examples, I just gave basic prompts and the AI executed: Balatro, but for coin flipping (all the design and ideas was Fable): https://play-flipside.netlify.app/ The best self-aware snake game: https://snake-stable-build.netlify.app/ An isochronic map using real data: https://isochronic-passage-chart.netlify.app/#syd
I've had access to Fable for a bit. A genuine jump in capability, I could feed it a 15 page design document for a project and it would work for 9+ hours and deliver terrific results.
But working with it is weird & weirder is coming
Lots of examples: https://open.substack.com/pub/oneusefulthing/p/what-it-feels-like-to-work-with-mythos?r=i5f7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
@emollick Thanks for testing it!
I've had access to Fable for a bit. A genuine jump in capability, I could feed it a 15 page design document for a project and it would work for 9+ hours and deliver terrific results.
But working with it is weird & weirder is coming
Lots of examples: https://open.substack.com/pub/oneusefulthing/p/what-it-feels-like-to-work-with-mythos?r=i5f7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
@arthur_spirling I generally get access to Big Labs models early. Some combination of being a researcher & being influential. I have never taken any money from any AI lab and they don’t see my posts in advance.
@emollick Can you share the prompts you used for the games? The snake one (http://snake-stable-build.netlify.app) is way more creative than anything I've seen from other models, so I'm curious how much of that was you vs the model
@arthur_spirling Nope. I don’t write about the vast majority of models I test. And no one has ever asked if I am going to do so.
@emollick Sorry if I missed it: could you say more about how you had early access to Fable?
@emollick 9 hours unsupervised and "terrific results" - that's the dream pitch. But "weird & weirder" is doing a lot of heavy lifting for a cliffhanger. What actually spooked you - the process or the output?
@emollick In case anyone is interested, had ChatGPT 5.5 Pro review the paper https://chatgpt.com/s/t_6a284f7783b4819191f96c0c5c5d638e. Briefly put: A solid paper, and the "flaws" are not major errors but elements that are typically modified during a review process. Note, I haven't read the paper fully myself.
@emollick Thank you. May I ask if the usual expectation is that you will write a post about the product?
@emollick Does it talk like Opus 4.8 ?
Definitely nothing to worry about.
This is only about an exponential improvement trajectory and RSI (recursive self-improvement – https://www.anthropic.com/institute/recursive-self-improvement).
What can go wrong?
I commented on RSI article, that there were clear if implicit warnings about dangerous capabilities, but this today is new level actually. https://darkomulej.substack.com/p/when-ai-builds-itself-a-warning-disguised
@emollick Is it $50/m output tokens terrific?
@emollick Is was hard not to feel something otherworldly while reading this & clicking thru images/examples. "Product release" no longer feels like the appropriate category.
@emollick Damn. I am not using it right. I had it format an excel for me and it still struggled.
Wanted it to match the exact format of a file I uploaded as an example ( image) and it still struggled.
@emollick the security cost hides in "9+ hours." approval used to be per action. a model that runs unattended that long collapses it to one decision at hour zero. you're not approving a task anymore, you're approving nine hours of judgment you never see.
@emollick ye similiar experience so far, its very careful model and verifies its own work a lot better than pervious model. its also very slow at least so far today
@emollick 9 hours means the model now has a longer attention span than I do. you can't watch a run that long, so you need something between babysitting it and finding out at dinner what it decided to build.
@emollick Thanks for sharing. Great to get a real world view of what’s possible. The marginal cost of software deployment just continues to plummet. Love the test ideas you used to put it through its paces
@emollick
@bribiotech @emollick Have you tried it in cowork? It's done amazingly well for PPT documents in cowork for me. Clear and design forward.
Ethan Mollick's early test of Anthropic's Mythos-class model shows it tackling a roughly 19-page design spec with almost no ongoing guidance, running for 9.5 hours while spinning up sub-agents, pulling real flight and rail data, writing code, and verifying its own outputs to deliver working software for statistical analysis.
I've had access to Fable for a bit. A genuine jump in capability, I could feed it a 15 page design document for a project and it would work for 9+ hours and deliver terrific results.
But working with it is weird & weirder is coming
Lots of examples: https://open.substack.com/pub/oneusefulthing/p/what-it-feels-like-to-work-with-mythos?r=i5f7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true