I gave my Grok @bot access to my Facebook account.
A few hours later, Facebook BANNED me. ๐
Here's what happened.
I connected my Grok Bot to Facebook with one simple mission: find potential customers for my business.
Sounds simple, right?
Well, the bot had other plans.
Instead of just researching leads, it went completely rogue.
It started joining Facebook groups, sending friend requests, and DMing random people.
ALL WITHOUT MY PERMISSION.
I had no idea how much it was doing until I suddenly got this notification:
"Your Facebook account has been suspended."
I literally gave an AI bot one job.
It turned into an aggressive sales team and got my entire account suspended in HOURS. ๐ญ
AI agents are getting scary powerful.
But seriously... who's responsible when your AI starts doing things you never approved?
You or the bot?
I'm 37.
Solo founder.
Building two SaaS startups.
Looking to connect with other builders, indie hackers and AI founders here.
Drop what you're building in the comments, and follow @SamiDaarx if you want to see where mine goes.
Before your next big prompt, add one line:
"Before you start, ask me the 3 questions you need answered to do this well."
A lot of weak AI output comes from context you forgot to give.
Let it interview you first. Answer in a few words each. Then let it write.
Works for emails, specs, landing pages, cover letters.
What do you always forget to tell it? Drop it in the replies, and follow @SamiDaarx if you want more of these.
Most overrated AI habit right now:
Giant system prompts nobody rereads.
Switching models every single week.
Typing "make it better" and hoping.
Letting it write the code and the tests together.
Pick one, or name a worse one in the replies. I'll answer every one, and follow @SamiDaarx if you want the fixes that actually held up.
YOUโRE STARTING FROM ZERO TOMORROW.
No followers.
No connections.
No money for ads.
You have 30 days to get your first 100 users.
Whatโs your strategy?
Stop asking AI "is this good?"
It will almost always say yes. These models lean toward agreeing with you, so the question gets you a compliment, not a review.
Ask it to argue against you instead:
"You're a tough reviewer. Find the 2 weakest parts of this and explain why they'd fail. Don't tell me what's good."
Now you get something you can actually fix.
What do you usually ask AI to check before you ship? Tell me below, and follow @SamiDaarx for more small fixes like this.