AI Is Not Making You Lazy. It Is Making You Quit Sooner.

Hey Everyone - Hope you had a good weekend. I have been heads down on the network I am building (more on that below) and reading a study that changed how I think about what AI is doing to us. This one connects to last month's piece on cognitive foreclosure, but with a sharper finding and a much sharper takeaway.

This week:

Let's dive in.

Researchers at UCLA, MIT, and Carnegie Mellon ran a simple experiment that has been sitting in my head for two weeks. They split participants into two groups. One group solved problems with an AI chatbot at their side. The other group worked alone. While the tool was on, the AI-assisted group performed better. That part was expected.

Then the researchers took the AI away.

The same people, on the same kinds of problems, now performing without the assistant. Their problem-solving accuracy dropped. That alone would be interesting. But the more important finding was about persistence. The AI-assisted group, once the tool was removed, became measurably less willing to keep trying. They gave up on hard problems faster than the control group, who had never had the assistant in the first place.

The study authors call this a "boiling frog" effect. I think that framing undersells what they actually found. What the experiment shows, for the first time with same-subject before-and-after data instead of correlations, is that AI use does not just atrophy a skill. It lowers the activation energy you are willing to spend the next time the problem is hard.

This matters because persistence is the trait nearly every form of independent work depends on.

Most weeks this newsletter is good because I sat with the draft longer than was comfortable. The version you read is the one I would not have arrived at if I had given up on draft three. The same is true of side projects, products, and almost every craft I admire. The work compounds because the person doing the work kept going through the quiet middle, the part where no machine is doing it for you.

The study is not saying AI makes you dumber. The study is saying AI makes you quit sooner. Those are different problems with different costs.

The first version is the one most of the conversation has been having. Cognitive offloading, losing your edge, letting the machine think. All of that work is real, and all of it is about cognitive surface area.

Persistence is something else. It is not a skill. It is the input that produces every other skill. If you cannot sit with a problem long enough to develop taste, you do not get taste. If you cannot keep building when the post does not pop, you do not get the audience. Persistence is upstream of almost everything humans still have over machines.

And the experiment is suggesting that AI use, in the cleanest study design so far, lowers your supply of it.

If you are doing creative work, the timing of your AI use matters more than the volume. A separate 2026 study found that participants who worked alone first, then consulted the chatbot, scored higher on critical thinking than those who used AI from the start. AI as a finisher, not as a starter. Do the hard part of the thinking before you bring the tool in, and the tool sharpens you. Bring it in first, and the tool replaces the muscle you needed to build.

If you are running a side project, treat your willingness to keep trying as a finite resource the tool is metering. Notice the post you almost did not publish. That moment is where the return on the project lives. Do not outsource the part of you that decided to keep going.

If you are a parent, the implication is harder. Adults losing persistence is recoverable. Children who never build the muscle may not get a second pass. The choice about when your kids meet AI is a thirty-year wager, and the new study is one more reason to make it slowly.

Persistence is the muscle. The tool is the remover. The work is the reminder that you still have it.

This week's building section is one that won't be for everyone, and that's fine. I've been quietly building a real network in Austin for about 4 months, and it's just started paying off in ways that surprised me.

I attend three to five networking events a week. That sounds like a lot, and it is. It a full month before any of it felt productive. I'd walk into rooms where I knew nobody and leave wondering what I had been doing. The early stretch is the part most people don’t like, which is exactly why most people don't get the back end of the curve.

The change happened slowly and then all at once. Now when I walk into a room I almost always know somebody. The people I know introduce me to people I don't know, and a non-trivial portion of those introductions directly support what I'm trying to build. I have a few sponsorships in the pipeline from these interactions alone. The newsletter, the Founders Feed, the next product, all of it gets easier the bigger and warmer the network gets.

The thing I didn't expect is the second-order effect. Once you know enough people, you start being able to make introductions for them. A friend mentions a problem and I realize I sat next to the exact person who could solve it last Tuesday. Making that connection costs me nothing. The other person remembers, the friend remembers, the network gets denser, and the flywheel spins a little faster. That part is what makes the work feel less transactional and more like being useful.

I want to underline the connection to this week's Signal. None of this works without persistence. The first month of going to events when you know nobody is the equivalent of writing the first ten newsletter posts that nobody reads. The whole strategy lives or dies on whether you keep showing up before the network has any reason to reward you. The output side, the introductions and the sponsorships and the real money, is the back half of a curve where the front half looked completely flat.

It's also turning into real money, which isn't the point but is worth saying out loud. The reason I keep mentioning revenue in the building section is that it's the cleanest signal that the durable work is actually working. Hedge bets aren't bets if they never produce.

Things I Learned:

Most networking advice is about how to walk into a room and meet people. The actual asset is downstream of that. It's becoming the person other people route through. The connector.

A connector is somebody who, when a problem comes up in their network, can almost always think of the right person to introduce. The introduction is the unit of value. Done consistently over months, it builds a reputation that compounds faster than almost anything else you can do for free.

The reason this matters now is the same reason this week's Signal does. AI compresses the value of information and execution. It does not compress the value of being the person who knows who knows.

Start this week. Pick three people in your network whose work you understand well enough to describe in a sentence: what they're working on, what they're looking for, what kind of help would actually move them forward. If you can't describe them at that level, you don't know them well enough to introduce them yet. That's your homework for the next event.

Then make one introduction. Email or text both parties at once. Two sentences on each person, what they're working on and why the connection is worth their time. The two of them take it from there.

The first few introductions feel awkward and most go nowhere. That's the same flat curve I mentioned in the building section, and it's exactly where AI use, per this week's Signal, would lower your willingness to keep going. The discipline is to keep making them when nobody has yet returned the favor.

This skill is durable regardless of what happens with AI. AI cannot generate trust between humans, only observe it. Whatever the next decade looks like, the person who is the connector node in three or four communities will be valuable in ways that have almost nothing to do with the model of the month.

Pick three people this week. Make one introduction. Then do it again next week.

When was the last time you stopped working on something hard right after the AI didn't give you a useful answer? Was it the AI's fault, or your own willingness to keep going?

Who are three people in your network you understand well enough to introduce thoughtfully today, and who is one person you'd need another conversation with before you could?

What introduction could you make this week that would cost you nothing and matter to both parties?

“Weekly AI Prompt: "I want you to help me audit my AI usage for persistence loss. Here are the most important hard problems I am working on right now: [list 3-5 projects, hobbies, decisions, or skill builds].

For each one, tell me:

Then tell me which of these projects has the highest risk of my willingness to keep trying being eroded by AI use, and what protocol I should adopt this week to protect it."