The machines score lowest on the thing you care about most

The story

Someone finally measured whether AI is good for people

For two years the argument about AI and human wellbeing ran on anecdote, and then a team put numbers on it. Gloo, working with Barna Group and Valkyrie Intelligence, built the Flourishing AI benchmark on top of the Global Flourishing Study and scored the leading models across seven dimensions of a life going well.

Read the averages as a shape rather than a scoreboard. Finances 81. Health 72. Relationships 67. Happiness 65. Character 58. Meaning 56. Faith 35.

The number to know: 35. Look at which end is which. These models are strong where a question has a procedure and weak where it has a person in it. Good with your money. Poor with your soul.

Pat Gelsinger, who ran Intel and is now executive chair at Gloo, made the case for measuring it at all in one line: we must measure AI "against the ultimate standard — human flourishing."

Why it matters for you: your people are already asking these machines the hard questions. Someone on your team is asking one right now about a marriage, or a job, or whether any of this still matters. They ask because it is there at eleven at night and never looks tired of them. That is exactly where the answers are weakest.

That is worth one deliberate act this week. Notice a question you were about to type into a model that was really a question about meaning, and take that one to a person instead.

The practice

Ask for one thing at a time

Most people write a prompt like a wish list. That is what breaks it. Two findings from this year's research are worth your Monday.

Meta tested models on prompts holding up to twelve rules. Most could not manage more than three. Give a model eight rules and it follows all of them about six percent of the time.

Researchers at EPFL, Apple and Mistral AI found the other half. They deleted the examples from a prompt, and accuracy rose from 74 to 84 percent.

So say what you want, and stop.

Weak, because it stacks nine instructions:

Write a note to my team about the reorg. Three paragraphs. Under 200 words.
No jargon. Warm but honest. Mention the timeline. Don't mention headcount.
End with a question. Match my voice from the sample below. Grade-8 reading level.

Strong, because it names one goal:

I need to tell my team about the reorg on Monday. They are worried about
their jobs and I cannot promise everyone keeps theirs. Help me say the
true thing without pretending it is smaller than it is.
Ask me what you need to know before you draft.

This week, take a prompt you use often and cut it down to three rules and one goal. Run both versions on the same task and keep the better answer.

Higher ground

The lowest scores are the parts of life nobody knows how to speed up. Character, meaning and faith form slowly, in company, with people who stay.

In 2017, before Clemson won the national championship, Dabo Swinney told his team the difference in the game "is gonna be love." He was talking about football. He was also naming the one thing no tool supplies. Whatever these machines get faster at, that part stays yours.

If I have prophetic powers, and understand all mysteries and all knowledge, but have not love, I am nothing.

1 Corinthians 13:2

The next step

If you want a straight read on where you actually stand with these tools, the AI Edge assessment takes about five minutes and tells you which strengths AI amplifies for you and which ones it quietly absorbs. Find your AI Edge.

And tell us how this landed. Hit reply with what helped and what you wish we had covered. We read every one.