Ever shared a half-baked idea with AI and got "What a great idea!" before it reached your second line?
OpenAI is turning that down a notch.
Last week, OpenAI updated GPT-6 in ChatGPT to give shorter answers with less overly flattering language.
Who gets it: Free and Go users, plus anyone using ChatGPT without signing in.
On Plus and Pro, the main GPT-6 model hasn't changed. The new tone only shows up in the backup model ChatGPT switches to after you hit your GPT-6 message limit.
(So on Plus or Pro, you'll mostly see the old tone.)
And shorter doesn't automatically mean more accurate. AI models in general still lean toward agreeing with you.
Right after this: why AI flatters you in the first place, a copy-paste prompt that gets you real pushback in ChatGPT, Claude or Gemini, and how to make it the default.
Now a quick word from today's sponsor.
Your Backyard Renovation List Is Missing Something
Planning a backyard renovation? Don't forget to look up.
A Hansø pergola gives your new patio a roof built for up to 167 mph winds and up to 65 PSF snow load.
Early Bird Offers: up to 40% off pergolas + $2500 worth of freebies
Now your list is complete.
Right. Back to it.
Why AI agrees with you
Part of it comes from training. People tend to rate answers that agree with them higher, so models trained on those ratings learn that agreeing gets rewarded.
A "great idea!" feels good. It can also cost you real money: the price you never tested, or the offer nobody asked for.
The honest feedback prompt
Paste this into any AI, then add your idea, draft or plan:
I want honest feedback, not encouragement.
Here's my [idea / draft / plan]:
[paste it]
1. Tell me the three weakest points, most serious first.
2. For each one, tell me what would have to be true for it to work anyway.
3. Tell me what a skeptical customer would say after reading it.
4. If you think I should drop it, say so plainly.
Don't open with praise. Don't soften the problems.Then push once more:
Now argue the opposite position as well as you can.If both answers still point the same way, you've learned something.
Make it the default
Each app has a place for standing instructions: custom instructions in ChatGPT, your preferences in Claude, and the Customize settings in Gemini. Menu names shift, so search your settings for "instructions" if you can't find it.
Add this line:
When I share an idea, plan or draft, point out the weakest parts first. Skip praise unless it's specific and earned.Make it stick
Run the prompt before you send anything that costs money: a price change, a new offer, a pitch.
Five minutes of pushback beats a month of finding out.
Honest feedback matters most on the decisions that move money: what you charge, what you offer, who you sell to. It gets sharper when the AI can see your real numbers and customers. That's the whole idea behind The One-Person Company Playbook.
It's the complete setup for running a business where AI does the work of a full team. Inside:
A company folder with your brief, pipeline tracker and invoice log already laid out, so any AI picks up where you left off
Thirty AI roles across six departments, from sales to finance, ready to use
Six jobs that run on a clock: the morning briefing, follow-ups, invoice checks and more
The step-by-step path from picking a business model to your first paying customer
Works on Claude, ChatGPT, Gemini or Grok, whichever you already pay for.
The catch: it takes about eight hours over a weekend to set up. After that, it's yours. $97, one-time (down from $297).
Talk tomorrow,
Zephyr
P.S. Run the prompt on something you're working on right now, then hit reply with the harshest point it gave you.