Claude Fable 5.1 is here — this one prompt shows what it can really do
When you buy through links on our articles, Future and its syndication partners may earn a commission.
Credit: Future
Claude Fable 5.1 arrived this week and it's Anthropic's latest high-end model, built for complex tasks that require sustained reasoning . Its biggest focus is agentic work, meaning it can plan, use tools, check its own progress and keep working through problems with less supervision.
With a 1-million-token context window, it's useful for huge codebases, lengthy documents and research-heavy projects. Anthropic says it delivers major gains in coding and long-running tasks
while reducing the cost of some agentic workloads through cheaper caching. The simplest way to think about Fable 5.1 is that it's designed to be given a project, not just a prompt.
For those new to Claude
who want to dive directly into Fable 5.1, here's the one prompt I suggest you try first.
Give Fable 5.1 a project
Credit: Getty Images
Fable 5.1 can handle more, which means you'll want to give it a sturdier and more meaningful prompt.
Here's a good prompt to see what Fable 5.1 can do: I'm building [describe a real project you're actually working on, in detail]. Here's my current approach: [paste your code, outline, or plan]. Tell me what's wrong with my thinking, what I'm missing, and then refactor or redesign the whole thing — but explain your reasoning at each decision point so I can learn from it.
While this prompt may seem deceptively simple, it makes sense for an AI that can do so much more than previous models. I especially like it because you're not starting with a blank slate. Instead, you're giving the AI something of yours that you've already started. That's always a recipe for a better and more thorough response.
Instead of asking Claude to create something from a clean set of instructions, you're handing it work that already exists, complete with your assumptions, mistakes and potentially questionable decisions. From there, Fable 5.1 has to figure out what you're trying to accomplish before deciding whether the way you're doing it actually makes sense.
And crucially, you're explicitly giving Claude permission to tell you that you're wrong. One of the least useful things an AI assistant can do is take a bad premise and enthusiastically help you execute it. This prompt asks Fable 5.1 to interrogate the premise first. Only then does it get to rebuild the project.
What the prompt is really testing
Credit: Shutterstock/jud-mackrill
There are several jobs hidden inside this prompt. First, Claude needs to understand the project and the approach you've already taken. Then it has to identify weaknesses that you may not have noticed yourself.
After that comes the harder part of deciding which problems actually matter and where the AI should begin.
A model could easily nitpick everything it sees and produce a huge list of suggestions without making the project meaningfully better. A useful response requires judgment about which decisions should stay, which should change and what those changes will affect elsewhere.
Finally, Claude has to redesign the work while keeping track of those decisions.
That's why I particularly like the final instruction to explain your reasoning at each decision point so I can learn from it.
With this prompt, you get to see what Claude knows and why it chose one path over another. There is an important caveat here. AI models don't provide a literal transcript of their private internal reasoning, and you shouldn't treat an explanation as a window into exactly what happened inside the model. What Claude can provide is a useful explanation or justification for the decisions it made
What should you give Fable 5.1?
If you're a developer, you could paste in code for something you've been building and explain what you're trying to accomplish. But you don't have to write code.
You could give Claude a business plan you've been struggling with, the outline for a novel that isn't quite working, a research project, a presentation, a home renovation plan or almost anything else where you've already made decisions and aren't sure they're the right ones.
In fact, I'd argue that the messier the project is, the more useful the test becomes.
A perfectly written brief gives the model a relatively easy job. Real projects aren't like that. They contain assumptions we haven't explained, ideas we've become attached to and decisions that made sense three weeks ago but don't anymore.
That's exactly what I want Fable 5.1 to find. When I put it to the test, I uploaded my entire manuscript for a book I'm working on. Then, I used this prompt without giving it any additional hints about what I thought was wrong with my work. Of course, that was the intentional part.
If I tell Claude where I think the problems are or what needs edits, I'm already steering the critique. I wanted to see whether Fable 5.1 would independently notice the same weaknesses — and, more importantly, whether it would find anything I had missed.
What surprised me
Credit: Shutterstock
I was surprised by how thorough Fable 5.1 was with its feedback and line edits. It offered suggestions without being overly editorial or trying to remove my voice at all.
One of my favorite things about Claude is that it doesn't just go ahead with edits the way ChatGPT or Gemini often will do. Claude questioned my assumptions, offered to change the structure for clarity and even found a problem with one of the chapters that I hadn't noticed.
From there, I let it outline what I would need to improve my manuscript and what other projects I could potentially develop from it.
But finding problems is only half the test. The second part of my prompt asks Fable 5.1 to "refactor or redesign the whole thing," which means it now has to turn its criticism into something better.
Because this was a manuscript and not a presentation, I didn't want to give it the reins in that aspect. Instead, I asked it to make a list instead of going in with all the changes. More importantly, Fable 5.1 explained why it suggested those changes.
Of course, I didn't agree with all of the changes, but that's what I like about using AI as a partner and not something that does the entire job for me .
Bottom line
If you're ready to ask AI for something beyond your typical query, Fable 5.1 is a great place to start. This is a great prompt — just be sure you don't waste your usage limits on a fake project. I mean, if you're going to try it, try it with something you're actually working on.
The reason Claude's prompt works is because you already know the material you're giving it. That puts you in a much better position to judge whether Fable 5.1's criticism is insightful or merely sounds convincing.
Take something you've genuinely been working on and give Claude enough context to understand both the project and what you're trying to achieve. Then let it disagree with you. This prompt asks Fable 5.1 to look at an answer we've already come up with and tell us why it might be wrong.
Give it a try and let me know in the comments what you think of Anthropic's latest model.
