Not every question belongs in your AI conversation

Writing “Chrome has all this?!” made me question my own legacy assumptions about the tool I use most: my AI assistant.

The Codex app, now part of ChatGPT, has evolved considerably over the past few months. The way I use it mostly hasn’t.

I take advantage of subagents, but most conversations still follow the same pattern: I ask something, it answers, I follow up, repeat.

While selecting text in a conversation, I noticed a new option: Ask in side chat.

Normally, getting clarification means adding another question to the main conversation. It gets answered, but the conversation becomes longer and I eventually have to point the assistant back to what we were originally doing.

Side chat gives that temporary question somewhere else to go.

Ask the question. Get the answer. Close it. Then it disappears.

Different assistants implement the idea differently. ChatGPT calls its temporary version Side, while Claude Code offers Side Chat and the /btw command.

The practical rule is simple:

• If the answer should change the work, ask in the main conversation.

• If you only need clarification before continuing, ask on the side, let it die there and keep the main conversation clean.

If you want help making better use of the AI tools you already have, without adding more complexity: let’s talk.