In Plain Sight - 09
In Plain Sight is a weekly blog from Rhys that we will publish every Thursday/Friday. Each edition will be focused on a thorny topic within investing, startups, learning and work and policy.
In Plain Sight is partially a personal attempt to think independently and to avoid leaning too much on AI for answers to our questions. Will we use AI to edit and polish the text? Yes. But more importantly, will we come up with all of the ideas and analysis? Yes.
Here follows Edition 08 - ''Smart' is not what it used to be'
------
Is the next generation getting less intelligent, or are we measuring the wrong thing?
Around 370 BC, Plato has Socrates warn that a new technology will ruin young minds. Writing, he says, will "create forgetfulness in the learners' souls". People will stop using their memories and rely on marks on a page. They will look wise without being wise.
More than two thousand years later, the worry is back, as it has been at several other points in history since Socrates' words of warning...
Over the past year, a claim has spread from government hearings to front pages: Gen Z is the first generation in modern history to be less cognitively capable than its parents. Test scores are down, attention spans are shorter and a new wave of AI tools threatens to do students' thinking for them.
We take this seriously, given our learning and work thesis. But we think the headline may be missing something. The question isn't only whether young people are getting less smart - importantly, it's whether what it means to be smart is changing and whether we're still measuring the old version.
It's worth starting with the strongest version of the worried view because this is where the news currently focuses.
Earlier this year, neuroscientist Jared Cooney Horvath told the US Senate Commerce Committee that Gen Z is the first generation in modern history to score lower on standardised tests than the one before it. He pointed the finger at screens in classrooms. The US spent more than $30bn putting laptops and tablets into schools in 2024 alone and, in his reading, learning got worse.
The longer-run data points the same way. Through most of the 20th century, IQ scores rose by around three points a decade, a pattern known as the Flynn effect. Then, in several rich countries, it stalled and reversed. A study of more than 736,000 Norwegian conscripts found scores peaked among men born in 1975 and fell after that. Crucially, the drop showed up between brothers in the same family, which rules out genetics and points to something in the environment (Bratsberg & Rogeberg, PNAS).
The FT's John Burn-Murdoch has shown reasoning and problem-solving scores falling across high-income countries since around 2012, alongside a steady rise in teenagers reporting trouble concentrating. His diagnosis is behavioural: we read less and we've swapped finite web pages for infinite, constantly refreshed feeds (via AEI).
And now there's AI. An MIT Media Lab study had students write essays with ChatGPT, with a search engine, or with nothing at all. The ChatGPT group showed the weakest brain engagement and struggled to recall what they had just written. The researchers called it "cognitive debt" (MIT Media Lab).
It's an uncomfortable picture. But it does raise a question: are we measuring a decline in intelligence, or a decline in one particular kind of it?
Intelligence is not a fixed thing
Let's head back to Socrates. He was partly right. Writing did weaken the extraordinary oral memory that let bards recite epics from start to finish. But nobody today would call the invention of writing a net loss for human intelligence. It gave us science, law, history and literature. It changed what a smart person needed to be good at. That pattern repeats.
James Flynn, whose name the IQ gains carry, never believed our grandparents were dim. His explanation for the rising scores was that modern schooling and office work trained people to see the world through "scientific spectacles": to classify, to reason about hypotheticals, to think in abstractions. Those are exactly the habits IQ tests reward. Scores climbed because the world started rewarding what the tests measured.
If that's why scores rose, it's worth asking whether the same logic explains why they're falling. The Norwegian brothers study tells us the cause is environmental. It doesn't tell us the environment is making people less capable in every sense. It may be training something the tests weren't built to see.
We've seen smaller versions of this before. Mental arithmetic faded once calculators arrived. A 2011 study in Science found that when people expect to be able to look something up, they remember less of the information itself, but get better at remembering where to find it (Sparrow, Liu & Wegner) - which feels more like a trade than a loss.
So perhaps the real question isn't whether young people are smart. It's which kind of smart they're being trained for and which kind the world will reward.
The simplest way we've found to frame the shift is this: intelligence is moving from storage to steering.
Storage smart is what you can hold in your own head. It's recalling facts and formulas, solving a well-defined problem on your own, keeping your focus on a single text for an hour and getting the right answer under timed, closed-book conditions (think back to our In Plain Sight about the time it takes to qualify for specific accreditations…). It's what most of our exams and most of our hiring proxies were designed to measure.
Steering smart is what you can direct. It's knowing what to ask and where to look, framing a messy problem before trying to solve it, pulling together many sources quickly, orchestrating tools and people and, above all, judging whether an answer is any good, whether it came from a colleague, a search engine or an AI model. A non-tech version of 'steering smart' is an open-book test, which is an increasingly common form of assessment, as is oral assessment.
Employers are already moving the goalposts. The World Economic Forum expects 39% of workers' core skills to change by 2030. The fastest-growing skills on its list are AI and big data, technological literacy, creative thinking, resilience and flexibility, and curiosity and lifelong learning (WEF Future of Jobs 2025).
Importantly, analytical thinking still tops the list of core skills, cited by 70% of employers. And job postings for generative AI roles ask for more cognitive skill (AEI, citing this analysis of 7.2m job postings). Steering doesn't replace thinking, but it does change where the thinking happens.
Picture a 16-year-old who struggles with long division in their head, but who can spend a weekend prompting, testing, debugging and shipping a working app. A standardised test sees a weak student. A hiring manager in 2030 might see something else… none of which means distraction is a new form of genius.
It would be easy to stop here and conclude that it's all fine and there's nothing to it - we don't think that's quite right either.
You can't steer what you don't understand. Judging whether an AI's answer is right takes knowledge and reasoning, the very things the test scores show slipping. A student who has never learned how an argument is built can't spot a weak one. Steering sits on top of storage rather than replacing it.
Cognitive debt compounds. If tools do the thinking during the learning phase, as the MIT study hints, people may never build the foundations they later need to steer with. The calculator didn't hurt people who already understood arithmetic. It's less clear what it does to people who never learned it.
"The definition is changing" can be a cop-out. Every struggling system can claim it's being measured on the wrong thing. If we believe steering matters, the burden is on us to measure it properly, not just to wave away the old measures.
Attention is a precondition, a pre-requisite. Sustained focus underpins almost every kind of smart, old or new. Feeds built to maximise engagement erode it and that feels undeniable.
The average may hide a split. Young people with strong foundations and good guidance may become exceptional steerers. Others may become passengers, handing over their judgement along with their homework. A falling average could conceal a widening gap.
So the choice isn't between this version of old smart and new smart - rather, it's whether we deliberately build the floor on which steering is built...
To conclude...
For educators and policymakers, we believe the job is twofold. First, protect the floor: reading, numeracy and sustained attention, taught deliberately, with far fewer screens during foundational learning. Second, start assessing steering too. Open-book, AI-allowed exams that grade judgement, problem framing and verification would tell us far more about readiness for work than another round of recall.
For employers, it's time to stop hiring on proxies for storage. Grades and recall-heavy tests say little about whether someone can frame a problem or catch a confident but wrong AI answer. Attention and judgement can be treated as skills to train, not fixed traits to screen for. We'd argue this is already happening.
We think there are important implications for parents too:
- If children ask AI everything, teach them to cross-check what it says.
- If children skim, make one long book at a time non-negotiable.
- If children build things with AI tools, ask them to explain why it works, not just that it works.
- If children can't focus, treat attention like fitness: something built through practice, not willpower alone.
Socrates worried that writing would leave people with the appearance of wisdom but not the substance. Today's worry is much the same, with the chatbot in place of the paper. He was right that something would be lost. He was wrong that nothing would be gained.
Maybe Gen Z isn't the first generation to be less smart. Maybe it's the first whose smarts our tests can't fully see and one that still needs the foundations those tests were built to check. Whether that's reassuring depends entirely on what we build next.
If you're building tools that help people learn to think with AI, rather than hand their thinking to it, we'd love to talk to you.
.png)
