Thariq floats killing plan mode in the harness
The shift+tab hotkey would go to adjusting effort levels instead.
Thariq said the team is thinking of killing plan mode and reusing the shift+tab hotkey to adjust effort levels instead. His reason is short. He does not think the models need plan mode anymore, though he asked plan mode diehards to tell him why he is wrong.
Matt Pocock's answer was to kill it, and let users plan how they want. His argument is not that planning is useless, it is that planning is the wrong thing for a harness to own. The way people plan will be custom for the task, org and person, he wrote, which makes it a userland concern rather than a built in mode.
Kent C. Dodds offered the data point rather than the argument. He said he has not used plan mode in months, and that with models as good as they are now you need two things, good and discoverable primitives, plus a short conversation about how to use those primitives to do what you need done.
What is actually being argued
Nobody in the thread said planning stopped mattering. The disagreement is about where it lives. A mode is a contract the tool maintains, with its own hotkey, its own state and its own behavior everyone inherits whether it fits their workflow or not. Pushing it to userland means the planning step becomes a prompt, a file or a small script each team writes, which is more work for the user and less for the maintainer.
That is also why the hotkey detail is the interesting part. Shift+tab is prime real estate in the interface, and the proposal is to spend it on effort levels rather than on a planning state. If effort is now the dial people reach for most, that says something about how the harness thinks users are steering these agents.
Worth being precise about the evidence here. This is three people on X, one of whom builds the tool, one of whom argues from design principle, and one of whom reports a personal habit over a period of months. It is not a survey, and Thariq is explicitly still collecting pushback from people who use plan mode heavily. If you are one of them, the thread is the place to say so.
For working engineers, the practical read is that plan mode's removal would not take planning away, it would take away a standard place for it. Teams who have built their planning into instructions or scripts lose nothing. Teams who lean on the mode because it is there would have to write down what they were relying on, which is the kind of exercise that usually reveals it was thinner than it looked.
Don't mandate a 'mode' and users AND maintainers will be happier
I've not used plan mode in months