Isn't that close to what instruct-tuned LLMs do when they are adviced to first think out aloud before jumping to answering? They are generating their own output

Isn't that close to what instruct-tuned LLMs do when they are adviced to first think out aloud before jumping to answering? They are generating their own output to their scratchpad, the attended context, to support their thinking.