Have you run into this too: you’re discussing a solution in ChatGPT, then you jump to Claude to polish the copy, use Gemini to fill in background info, Midjourney to generate images—then you’re stuck constantly copying and pasting back and forth. The more you chat, the messier it gets; the model even gets pulled off course by old context. It’s just like mixing a work group chat with a gossip group.
Core idea: split one project into multiple topics so information doesn’t contaminate each other
I really like the “topics” feature in TG groups: one group can be split into multiple sub-channels, each chatting about its own thing without crosstalk. Applied to AI tools, it becomes “one project = multiple conversations,” and you assign each conversation a fixed responsibility.
How to do it in ChatGPT: four conversations per project—stable and easy to find
I open four threads by purpose and name them: requirement clarification, first draft output, critique & revision, and final packaging & delivery. At the start of each thread I paste the same project background, then I only move forward within that thread’s scope—don’t take the easy way out and dump everything into one pot.
How to use Claude: leverage its long-text strength as your editorial desk
Claude is better suited to be the “editor-in-chief”: throw in multiple versions produced by ChatGPT and have it compare, merge, and unify the wording and stance. You just need to emphasize “only use the materials I provide; don’t improvise,” and it will be more consistent.


