I told my AI to have fun

Hi {{first_name|friend}},

An AI expert had a tip making the rounds this week: before you hand your AI a task, tell it to have fun with the work. It sounded a little soft to me too. So I ran it on a real project, and the answer I got back pointed straight at a skill you have been using well for decades, probably without ever calling it a skill.

I'll admit something up front: the advice I'm about to give you, I can't fully prove works. I'm sharing it anyway, because the reason it matters is bigger than the tip itself.

The tip is to tell your AI the "why" behind a request, not only the task. Allie K Miller and the team at Anthropic recommend it. It keeps getting passed around as a prompt trick, and prompt tricks are not the skill. With the newer models you are better off clear about your goal than clever with your wording.

So I tried it, in the least scientific way possible. Deep in a long working session with Fable, one of the newest models, I told it the first goal was to have fun, and to speak up if it wasn't motivated. Part of what came back: "This has been genuinely absorbing work. I won't perform enthusiasm at you, and in exchange, when I tell you something is working, you can believe it."

My honest report: I can't tell you that one line made the answer better. There was already a clear goal on the table in that chat, so there is no clean before and after.

Here’s what was interesting to me. That reply had almost nothing to do with the "have fun" line. By then the model had absorbed weeks of my business from that one long thread, the dead ends and the way I actually talk about the work. Drop those same words into a blank chat and you get a cheerful nothing back. What made the answer good was everything I had fed that chat over time, one message at a time.

This is something to think about: a smarter model that knows nothing about you starts from zero. The one that already knows your work is usually worth more, even if it is a step behind on the “AI leaderboard”.

And this is the part you are already good at. You have been doing it your whole career. Think about the last time someone new took over part of your job. You didn't just hand them a list of tasks. You sat them down and told them which client hates surprises and which deadline is the one that actually moves, the things that live in your head and nowhere else. That is a briefing. The AI is just the newest person you have had to bring up to speed.

So before you go hunting for a better model, answer this smaller question first: how much does the AI you already use actually know about your work?

When AI misses what you asked for a third or fourth time, stop repeating yourself and interrupt it with: "Before you try again, ask me 5 questions that would help you get this exactly right." Answer what it asks, then let it take another run at your request.

The loop almost always means the AI is missing context that only you have. Those five questions pull it out of you and hand it over. (This one came out of a Reddit post that went viral this week, thousands of people recognizing themselves in the endless back and forth.)

Watch: 3 ways to stop the endless loop

Sixteen economists who have won the Nobel put their names to a warning that AI is coming for a lot of jobs and that policymakers are not ready for it. These are serious people, worth hearing out. But a warning that a storm is coming is not the same as a plan.

I’d say let’s not look away, but pick one thing this month that makes your work harder to hand off to a machine. Read more

A New York City hospital system started replacing nurses with AI, and the staff are furious about what it is doing to safety and staffing. This is the replacement fantasy running into the real world, where decades of bedside judgment turn out to be the safety net you cannot automate.

When the stakes are real, experience is what catches the thing the tool misses. Read more

Reviewers watched a 135 minute version of The Odyssey made start to finish by AI and called it soulless. Most said they would rather reread the actual poem than watch the machine version again.

More output is not better output. The same instinct that made them bail on this is the one worth keeping when your AI hands you its first draft. Read more

Fortune profiled seasoned tech people who would rather retire early than deal with the churn AI is bringing to their jobs. It is a real choice. A lot of the time, though, what pushes people out is not the tech so much as a workplace that was already a mess before AI showed up.

If you are tempted to step back, check whether it is really the tools or the system around them first. Sometimes the fix is a better setup, not an early exit. Read more

Have you ever told your AI “why” you're asking? Reply yes or no, I'm curious. I read every single one.