Matt Pocock on X: "@arthur_alb1 One weird thing about 'plan mode' is that it convinces everyone they need a plan
A plan is only useful if you need a summary of the decisions made OR if you're handing off to another session.
If not, just grill + implement"
One weird thing about 'plan mode' is that it convinces everyone they need a plan
A plan is only useful if you need a summary of the decisions made OR if you're handing off to another session.
If not, just grill + implement
Matt, I use grill-with-docs after creating a plan to improve the plan. I now realize this is not how you do it though.
How exactly you do it? You start a brand new session with no context and no plan and create a plan after grilling?
One weird thing about 'plan mode' is that it convinces everyone they need a plan
A plan is only useful if you need a summary of the decisions made OR if you're handing off to another session.
If not, just grill + implement
In my experience, a plan nearly always gives me the quality I want. Superpowers (brainstorming) + grill-me producing a design and then an implementation plan is very time consuming and burns lots of tokens initially, but I get what I want sooner (less rework).
To save tokens, I enjoy planning with Opus and then provide the plan to a chepear model, as in OpenCode Go
So I activate plan mode and, after Opus writes the first draft, I call grill-with-docs to improve it. Should I call this skill at the start of plan mode instead?