Two trends are moving in opposite directions. AI is increasing in capability, and at the same time, humans are atrophying and becoming more apathetic.

The first trend shows up clearly in AI benchmarks. The second shows up in standardized test scores, and anecdotally, I think any educator at any level can attest to it to some degree.

I want to be careful here. AI didn’t cause all of the decline I describe below. But AI does make it easier than ever to skip the work that forms our human knowledge and skill, and that’s what concerns me.

AI keeps getting more capable

In 2025, the research group METR reported that the length of tasks AI agents can complete on their own (measured by how long those tasks take skilled people) had been doubling roughly every seven months for six years. Their updated estimates suggest the pace has picked up since 2024.

Stanford’s 2026 AI Index found the same thing from a different direction. On one widely used coding benchmark, the best scores went from about 60% to nearly 100% in a single year.

Note that benchmarks don’t capture everything, and a high score doesn’t guarantee good work in the real world. But the direction is clear.

Human skills are slipping

The direction is clear on the human side, too. It just points the other way.

In 2024, the average 12th-grade scores in both math and reading on the National Assessment of Educational Progress (NAEP) were the lowest ever reported on each assessment. It isn’t only students. The average literacy and numeracy scores of U.S. adults fell between 2017 and 2023, and the share of adults at the lowest literacy levels rose from 19% to 28%.

The pattern shows up in reasoning, too. For most of the 20th century, IQ scores rose from one generation to the next (the Flynn effect). That rise has stalled or reversed in several countries. In Norway, scores rose for men born between 1962 and 1975 and then declined, and comparisons between brothers showed that the causes are environmental, not genetic. A study of nearly 400,000 U.S. adults found declining scores from 2006 to 2018 in three of the four areas it tested, though the authors caution that the decline may reflect test-taking or motivation rather than ability.

None of these trends started with ChatGPT. Phones and the pandemic likely play a big part. But AI adds something new. It can do the very work that builds a skill, and do it well enough that no one notices right away.

The early evidence points that way. In one study of high school math students, those with unrestricted access to GPT-4 scored 17% worse on exams once the AI was taken away. And at one large university, the share of A grades in courses where AI can do much of the graded work (e.g., take-home writing and coding) rose by 13 percentage points after ChatGPT’s release, about 30% above the 2022 baseline. The increases were largest where homework counted most, consistent with AI substituting for student work. Grades are going up while the skills behind them may not be.

Apathy is harder to measure

Atrophy shows up in test scores. Apathy is harder to pin down, but teachers see it.

In a 2025 survey, teens reported feeling confident and motivated in their core classes, while their teachers saw much less motivation than the students reported, pointing to cellphones and the pandemic, among other causes. At 22 public research universities, students’ study time in and out of class has kept declining since the pandemic.

Apathy and AI feed each other. The less we care about the work, the more tempting it is to hand it off. And the more we hand off, the less there is to care about.

Why this matters for Christians

As Christ works through us, we exist to love and serve our neighbors and to steward God’s creation. That calling requires that we not delegate control to increasingly capable technologies.

We can always love and serve our neighbors, even in a hypothetical reality with AI “overlords.” God’s people have served their neighbors under foreign rulers before. In exile, they were told to seek the welfare of the city where God had sent them (Jeremiah 29:7).

But Genesis is clear about God’s intent. He made humans in His image and gave them dominion over the earth (Genesis 1:26–28), and He put the man in the garden “to work it and keep it” (Genesis 2:15). We don’t do this on our own. Christ and the Holy Spirit work in and through us (Galatians 2:20, Philippians 2:13).

Stewarding incredibly capable and complex technologies like foundation models is challenging. It’s nonetheless part of God’s calling for us.

Apathy isn’t a new temptation, either. Proverbs describes the field of the sluggard, overgrown with thorns because no one tended it (Proverbs 24:30–34). In Jesus’ parable of the talents, the servant who buried what he’d been entrusted with is called slothful (Matthew 25:14–30). Atrophy, in other words, is what happens to gifts we stop using.

Note that none of this is about earning God’s favor. Christ is faithful even when we aren’t (2 Timothy 2:13), and when we fail, we’re forgiven. Resisting atrophy and apathy is a response to his faithfulness. We keep our minds and skills in shape because our neighbors need them.

What resisting looks like

The Stewardship question of the framework covers much of this. A few practices matter most to me:

• Do the work that forms you, especially while you’re learning (e.g., write the first draft yourself, work the practice problem before checking it).
• Stay competent enough in your field to check AI’s work. If you can’t recognize a wrong answer, you can’t steward the technology that produced it.
• Keep caring about the result, even when AI did most of the work. Ask what it possibly got wrong and whether it actually serves the people it’s for.
• Pay attention to what AI actually does on your behalf.

As AI agents continue to do more work on our behalf, I am concerned with several potential consequences. One of those, I elaborate on in AI Misalignment.