Claude is a (ridiculously) genius with the lights off.
These 7 settings switches turn them on:
- The model dropdown.
Bottom-right of the box. Most people never touch it.
Switch free model to Opus 4.8 (paid, but worth it).
Same screen, completely different brain.
- Effort: High.
One click, right next to the model.
It tells Claude to actually think instead of blurting.
Same question, but better answer. Free upgrade.
- Thinking, turned ON.
The toggle people leave off because they don't know what it does. It's Claude reasoning before it answers, not after you complain.
The single highest-leverage switch on this page.
- Connectors.
Customize → Connectors.
Claude reads your Gmail, your calendar, your notes.
It stops being a chatbot & starts being something that knows you. my favourite connectors are Gamma, Granola, Gmail & Calendar.
- Skills.
Customize → Skills.
Save a prompt with /skill-creator.
Type /yourskill forever.
Stop re-explaining yourself every single chat.
A skill ends that. Build it once, ship forever.
- The Cowork tab.
Claude desktop app, In between "Chat" & "Code."
The browser hands you text to copy-paste.
Cowork builds the real file - the actual PDF, Excel - and drops it in your folder. Not a copy of the thing.
- Bypass permissions.
In Claude Code (the scary one).
Everyone leaves it off. All it means is: Go. I trust you.
It stops Claude asking permission at every step so it can actually finish the job.
Off = it asks 40 times. On = it's done.
The ones who open Claude right now and flip all 7 just passed everybody who didn't.
Want the exact setup I use, with exact Claude skills?
It's right here → http://how-to-ai.guide.
Reply to the welcome email. Don't pay anything.
Open the Notion link in the second email.
Go to the "setup" folder and grab the checklist.
♻️ Repost this so your network stops using a genius with the lights off.