A new dial showed up, and I have spent more time thinking about it than most new dials deserve. It is called effort, and it controls something you could not control until recently: how deeply the machine thinks before it answers.
No new model, no separate plan. One model, and you turn how eagerly it spends its thinking. And since the latest Opus, that dial sits on high by default.
What it is, without the datasheet
You used to pick a model and take what you got. If you wanted deeper reasoning, you reached for a heavier, slower, pricier model. Effort lifts that choice out of the model name and puts it on a slider.
Low means fast and lean, no fuss. Good for hauling work. High means take your time, think it through, even when that costs more. And in between sits a middle setting for when you are not sure what you need.
The nice part is that it is not a hard ceiling. Set it low and hand over a difficult problem anyway, and the machine still thinks. Just less than it would with the dial wide open. It is a posture, not a cap.
The settings, as a feeling
I translated it into work I recognise.
Low is for moving things. Converting a file, cleaning up a list, a small helper that walks through a folder and comes back with what it found. Work where the answer is basically settled and only needs doing.
High is for deciding things. The shape of a piece that could still go anywhere. Code where the solution is not known in advance. A choice where I genuinely want some thought spent on the angle, not just on the first thing that works.
Above that sits room for the longer jobs. The kind of work that runs for half an hour or more, where the machine has to keep thinking as it goes. I rarely need it, but it is there when I do.
What actually changed
The shift is not in the dial itself. It is in where the dial starts.
It now starts on high. Everywhere. In the tools I work with every day, not buried in some setting. You used to have to switch good thinking on yourself. You began low and turned it up when the result disappointed. Now it is the other way around. You begin at thinking, and you step down on purpose when the job allows it.
That sounds like a detail and it changes the tone. The default assumption is no longer “do it fast” but “do it well”. Speed is something you now choose, not something handed to you.
Where I set the dial myself
Here my own distinction comes back, the one that keeps returning across this whole series. The difference between hauling and thinking.
The speed-up I feel every day is in the hauling. The moving, the building, the executing. There I set the dial low, because faster is pure gain. Nobody benefits from a helper that spends ten minutes brooding over a file name.
But I do not let the thinking speed up. For the structure of a piece, for a choice that matters, I open it up and accept the slower answer. Because that is exactly where I do not want the first thing that works. There I want some doubt to have happened.
And that is what the dial really gives me. Not the speed and not the depth on their own, but the question it puts to me every time. Is this a hauling job or a thinking job. The moment I ask that, I usually already know what to do.
What I take from it
I think putting the default on high is a good call. It says: assume thought, not haste. For most people who will never touch the dial, that is exactly the right setting.
But the real choice stays mine. When something is worth the slower answer, and when fast is good enough, no slider decides. That decision is still the work. It has only become a little more visible, and I am glad of that.
This piece belongs to an ongoing line about AI as an operator. Read also how I left WordPress and the assistant that executes.