Practical AI for people who still want to think. Use AI well—without outsourcing your judgment.
Joined August 2023
- Tweets1.1K
- Following158
- Followers41
- Likes139
Pinned Tweet
5 rules so AI makes me sharper, not lazier:
1. Decide what I think first, then ask.
2. Anything I'll act on gets a source I can click.
3. Ask it to argue against me.
4. Same question, fresh chat. If the answers differ, nobody knows.
5. Keep one skill I do without it, on purpose.
keepyourhead.co
A meeting used to just end. Now a lot of them leave a transcript that can't tell a decision from someone thinking out loud, and anyone with access can search it later.
If there's a notetaker bot in the participant list, give that a second's thought before you speak.
When rival companies all say the same thing on the same day, you have learned something real. You've learned about their incentives.
You haven't learned whether they're right yet. They might be. That's a separate question, and it needs checking on its own.
Google's AI box answers most searches before you click anything, in the same calm voice whether it's reciting a fact or making one up.
You don't need to check all of it. You need to check one part, and know when to skip the box entirely.
Then open the source it cites and find that detail on the page. The usual failures are an old page describing something that has since changed, or a page that never said it at all. Both are obvious once you're looking at the real thing.
Paste this when an AI meeting summary says you agreed to something:
"I'm [name]. From this transcript, list every task or decision assigned to me, and quote the exact line behind each one. If there isn't a line, say so."
Then find each quote yourself. It can make those up too.
Lazy version: you google something, read the AI box at the top, and repeat the number it gave you.
Better: click the small source link under it and find that number on the actual page. Check the date too. The summary won't tell you a rule changed since that page was written.
If a well known face turns up in a video offering you money, a giveaway, a trading platform, it's a scam. Not probably. It is.
No amount of frame by frame analysis beats that one rule, and the fake only has to hold up for the three seconds you give it.
Lazy version: a dramatic video lands in your feed, so you study the hands, the blinking, the way the hair moves. You find nothing, so you decide it's real.
Better: find who posted it first, and check whether anyone else is covering it. No visible flaws just means a newer model.
Paste this at the end of any question you'd actually act on:
"If you are not confident about a specific fact, write unverified next to it instead of guessing."
It won't make the model self aware. It makes hedging an allowed answer, which the usual format quietly punishes.
A lease, a job offer, a freelance agreement. Most people sign having understood about half, because the alternative is paying a lawyer a few hundred to read a rental contract. AI closes part of that gap. Here's how to use it without getting burned.
Then ask for three things. Plain English on what each side must do. The five clauses most likely to cost you money or freedom. And what's missing that a standard agreement usually has. That last one matters most. Contracts hurt you through the clause that isn't there.
Privacy advice about AI is mostly about protecting yourself. But look at what you actually paste in. An email thread is what your colleagues wrote. A message you want help replying to is your friend's disclosure, not yours. You can decide about your own data. Not theirs.
A fake recruiter can write flawless English and put a face on a video call. Both are cheap now.
What's still hard to fake is a company website you found yourself, instead of one they sent you a link to. Start there, before you judge how professional anyone sounded.
Apple's new camera mode signs a photo the instant you shoot it, so you can prove nobody edited it afterwards. Real thing, genuinely useful.
It proves a sensor saw that light. Point it at a monitor showing a fake and you get a signed, verified photo of a fake.
3am, the baby feels hot, and a chatbot is the only thing awake. Fair.
Ask it what to measure and which answers mean call now. Not what it is.
Under 3 months, 38°C (100.4°F) or higher means call the doctor now, not chat.
Save your local nurse line today, in daylight.