Everyone wants to know the same thing this month: what can Fable 5 actually do?
The honest answer is: more than any model before it, and the gap grows the longer and harder the task. Anthropic’s own framing is that the longer and more complex the work, the bigger Fable 5’s lead. It ran a codebase-wide migration for one company in a single day that would have taken a team over two months by hand. It can work for days at a stretch, planning across stages and checking its own output. It is, by a distance, the most capable tool our field has ever been handed.
And right on schedule, the other stories are piling up too: projects that went off the rails before they were finished. Fast, confident, and pointed in the wrong direction.
My friend Soumyo Mukherjee framed it better than I could. Using a model like this without real experience, he said, is like handing someone who just got their license the keys to a Formula 1 car.
Sit with that, because it inverts the thing almost everyone gets wrong.
More power does not mean less skill
The comforting story about AI is that a more powerful tool means you need less expertise. Formula 1 says the opposite. A family sedan forgives a beginner. An F1 car puts them in the wall on the first corner, precisely because it is so fast and so responsive to every input. All that capability doesn’t lower the skill required to handle it. It raises it.
Fable 5 is that car. It amplifies whatever you bring to it — including your mistakes — and it does it at speed.
“Just got their license” is the part I would underline. It doesn’t mean the new driver is foolish. They’re qualified. Legally cleared. Perfectly able to operate the vehicle in an empty parking lot. But the distance between holding a license and driving an F1 car is vast. It is the same distance between “I can prompt the model” and “I can ship a system my business depends on with it.” The license is the easy part. The reflexes are earned slowly, the hard way.
Why it goes off the rails
Here is the mechanism underneath the horror stories, and it is worth understanding, because it isn’t a flaw you can patch out.
The very thing that makes Fable 5 remarkable — long, autonomous, multi-day runs — is the same thing that makes its failures bigger and later. An F1 car punishes a small mistake instantly and totally, not gradually. A model running fast and on its own for two days does the same: a wrong assumption in the first hour doesn’t drift gently, it compounds, and every confident step after it is built on the error. You don’t notice at hour one. You notice at day two, when the whole thing is polished, plausible, and wrong.
And the model checking its own work is not the safety net it sounds like. It can verify that the code runs. It cannot tell you that the entire direction was wrong. That is a judgment about the destination, not the driving, and it is exactly the judgment a new license doesn’t come with.
What the experienced driver actually knows
Watch a great driver and the skill you notice isn’t going fast. Anyone can go fast in a straight line. The skill is knowing where the limit is, feeling the grip start to go a beat before the spin, and braking at the right point in the corner. It is the stuff that only lives in someone who has done it ten thousand times.
That is what a senior engineer brings to a model like this. Not faster typing. The taste to feel a build understeering toward the wall before it becomes a wreck — to catch the confident, wrong architectural turn on day one instead of paying to rebuild it on day thirty. The model supplies the speed. Experience supplies the part that keeps the speed from killing you.
Top-end model, top-end people
So when people ask what Fable 5 can do, I think it is the wrong question. The better one is: who is driving while it does it?
This is the whole reason we work the way we do. We put the most capable tools in the hands of the most experienced people, and we keep it that way on purpose — senior engineers with decades of scar tissue steering the model, catching what it gets wrong, and owning the result. Not juniors handed the keys and told to floor it. It is the same discipline I described when we rebuilt our own site in a weekend: the tool removed the friction, and thirty years of experience supplied the quality bar.
It is also the same point I keep landing on. A camera doesn’t make you a photographer. An IDE doesn’t make you a developer. And a Formula 1 car doesn’t make you a driver. The machine got faster this year. That didn’t make the driver matter less. It made the driver matter more than ever.
So that is the bet: the most capable model there is, in the most experienced hands we can put on it. And right now, while everyone is kicking the tires on Fable 5, we want to show you what that actually looks like. Send us an idea — something fun, something weird, the app you would love to see exist but never got around to building. We will point Fable 5 at it, senior hands on the wheel, and show you what comes out. Throw us your idea.
And if what is actually keeping you up at night is a project that already went off the rails, we can bring that one back too. Send the signal.