ChatGPT, Claude, and the rest all work the same way underneath: the harder you let them think, the slower and more expensive the answer. Ask a light question on the heavy setting and you wait for nothing. Ask a heavy question on the light setting and you get a shallow answer that sounds right.
The fix isn't a better AI. It's knowing which setting fits which task — before you type.
In ChatGPT: the model picker at the top of the chat. Higher models and "thinking" options are the heavy settings; the fast/auto options are the light ones.
In Claude: the model menu in the chat box. Opus with extended thinking is the heavy setting; Sonnet is the workhorse; Haiku is quick polish.
That's it. The dial was there the whole time — it just never introduced itself.
Planning your daughter's wedding budget is Tier 1. Drafting the invitations is Tier 2. Fixing a typo in them is Tier 3. Same wedding.
An AI that already knows your project is worth more than a "smarter" one starting cold. Re-explaining everything is the most expensive step nobody counts.
When unsure, try the lighter tier first. A weak answer costs you seconds. Defaulting to the heavy setting for everything costs you all year.
Set the dial before the task starts, and let the task end before you change tools. Switching mid-stream throws away the context you just built.
A routing card built for your role, your business, or your household — which tasks go where, on the tools you already have. That's a conversation.
Talk to Hohenja