University Might Be Move Two
For a very specific kind of school leaver, the better first move may be a year or two close to real work.
My wife and I are close enough to this decision that it no longer feels theoretical. Our eldest is two years from the end of school. If I were weighing this for a kid who was already showing unusual agency and a real appetite for building, I would not treat university as the automatic first move.
For a very specific kind of school leaver, I think university may now be better as move two than move one. The reason is simple: building first has started to look more reversible than it used to, and the upside of buying real reps early has gone up.
That sounds overstated because AI is already everywhere. But I think AI is in the strange phase where it is overexposed in headlines and still underappreciated in its actual consequences. The hype is loud. The leverage is still underpriced.
New Zealand's Three-Year Window was about what AI changes for countries and companies. This post is about what it changes for a parent looking at an eighteen-year-old and deciding how much structure, risk, and freedom to buy next.
By founder-shaped I do not mean a bright kid who likes the idea of startups in theory. I mean someone with unusual agency, tolerance for embarrassment, curiosity about customers, and a habit of making things happen without being asked. This is a narrow slice.
Why this moment is different
When I left school in 2000, the founder path for someone like me was much narrower. I failed my final year badly enough that university was no longer really an option. I taught myself to code, eventually got to Xero, later built Akahu, and now run Dolla from Matakana with zero employees. Back then you had to become the translation layer between an idea and working software. That took years.
That translation layer has clearly weakened.
A motivated school leaver can now use AI to prototype product, write code, research a market, draft sales material, automate admin, and compress the distance between an idea and a first customer to a degree that barely existed before. Not perfectly, and not without supervision from reality, but far enough that the economics of starting have changed.
That does not mean AI thinks for you. It does not remove the hard parts. You still need judgment, taste, domain understanding, and the willingness to put something ugly in front of a real user. But the leverage one person can now get from the tools is unlike anything most founders in history had access to at eighteen.
For most of history, a founder-shaped kid left school with three broad options. Get credentials first. Get hired first. Or spend years getting technically competent enough to build anything yourself. In 2026 there is a fourth option. Learn by building immediately, with AI handling more of the execution layer than a young founder could previously afford in people or in time.
That matters because reps are what compound. A school leaver can now spend two or three years learning product, sales, operations, and software by actually doing them, not by waiting to be allowed near them.
The window is temporary
Leverage alone is not the whole story. The bigger point is that the market has not fully repriced around this yet.
In New Zealand's Three-Year Window I argued that each technology shift has a short period when the new architecture is available before the rest of the market reorganises around it. The internet had one. Cloud had one. Mobile had one. AI has one now. If that pattern holds, the AI-native window opened around 2023 and starts narrowing around 2027 or 2028.
That is exactly why this is such a good moment for a parent to think harder than the default. The tools are real. The institutions are still adapting.
Universities still do some things unusually well: concentrated peer group, technical depth, research habits, and signal. Good companies can still be excellent training too, especially close to a strong operator. But most universities are still organised around a world where humans remain the main execution layer, and most companies are still bolting AI onto systems, processes, and org charts designed for humans. Junior jobs are still being described as if the old ladder is intact. Most of the economy is acting like the tools are an enhancement to the old model rather than the beginning of a different one.
That gap will not stay open forever. Existing companies will catch up. Universities will rewrite courses. Compliance teams will lock more down. Best practices will harden. The weird advantage right now is that a young person can still learn the tools in the wild, before the rest of the market has fully standardised how they are allowed to be used.
Why school leavers are unusually well placed
This is the part older people often miss. School leavers are unusually well placed to use this window.
They have almost none of the baggage. No mortgage. No management title. No lifestyle calibrated to a salary. No ten years spent mastering workflows that are now being rewritten. If they are wrong at twenty-one, that is not nothing. Two bad years can cost confidence, income, and momentum. But the switching costs are still usually lower there than they are at thirty-five.
That makes this one of the rare moments where the risky-looking move is also the two-way door. Spending two years buying real reps is cheaper at eighteen than it is at thirty. A founder-shaped kid can try, fail, recalibrate, go to university later, join a company later, or do both with much better information than they had at school.
The bet is not equally available
I should say the obvious thing plainly before going any further. This path is not equally available to every kid. It is much easier to call something a two-way door if the family can absorb a miss, offer housing, or help get them in the room with serious adults. For some kids, university or a job is not just convention. It is real insurance. My argument is not universal. It is a higher-expected-value sequencing bet for a narrow slice, and that slice is narrower if the family does not have runway to stand behind the bet.
If you are a parent without that runway, the honest answer is that university or a job with a stable employer may be the better opening move for your kid, even if they are wired this way. You can still give them the tools and the mindset to build alongside it. The "two-year bet" version of this argument is not the only path. It is one path, for a narrower slice again.
If you do make the bet, treat it like a bet. Timebox it. Decide who is paying for what. Set a review date. If twelve months pass with no shipped work, no serious adult willing to vouch for them, and no evidence of real customers or operational responsibility, stop calling it exploration and change the plan.
Walking back is not starting from zero
Walking back through the two-way door does not put them at zero, provided the two years produced something legible. Shipped work. Real users. A reference from an operator. A little revenue, or a measurable improvement inside a business. With those artefacts they arrive at university or into the workforce with a working feel for the technology that is reshaping both, not as a normal beginner.
Without those artefacts, the walk-back is into the same entry-level market everyone else is walking into, only later. The two-way door only works if there is something to walk back with. That is what makes the legibility test non-negotiable.
They know how to frame messy work precisely enough for AI to act on it. They know where model output breaks. They know what still needs human judgment, taste, and verification. They know how much leverage a single person can get from the tools when the problem is real rather than hypothetical.
That matters even if they never found a company. A school leaver who has spent two years using AI against real work, real customers, or real operational mess is ahead of someone who spent the same two years preparing only for the old execution model. The two-way door is not just reversible. The fallback is stronger.
Not uniformly. The fallback is strongest for commerce, operations, and customer-facing roles in companies that care how someone actually thinks. It is weaker where the gate is formal accreditation, technical depth, or a credential a hiring system cannot bypass. The argument is a sequencing bet for a narrow slice, not a universal substitute for a degree.
There is another advantage too. A school leaver can treat AI as the default execution layer from day one. They do not have to unlearn a decade of habits built around software humans had to drive manually. They can learn the tools in their native environment, not as a sidebar or a copilot, but as part of how the work gets done.
That does not mean young people automatically understand these tools better. Most do not. Age is not the advantage. Clean slate is.
Why university or incumbents can be the wrong opening move
This is where I part company with the default advice.
If you spend the next few years inside an existing company, you will probably learn useful things. You may learn sales, operations, industry detail, and how real customers behave. That can be valuable. But in most incumbents you will not learn the tools in their cleanest form. You will learn them inside a permissioned, retrofitted environment.
That is not because incumbents are stupid. It is because they have existing revenue, existing systems, and real downside. They are trying to fit new tools into software and workflows built for humans, which is the same retrofit problem I wrote about in Is Your Software Scaffolding?. They have compliance concerns, approval layers, security reviews, and a thousand reasons not to let the youngest person in the building rewire the place with frontier models. In many cases they are right to be cautious. The mess inside real businesses is also part of the education, which is why working close to a strong operator can still be excellent training. My argument is against low-agency, permissioned environments, not against all existing companies.
The same problem shows up in education, although the best universities still offer things that are hard to replace: peer group, formal methods, technical depth, and research environments. If part of the old value of a computer science degree was training you to be the translation layer between human intent and machine execution, that part now looks different. As I argued in Who Becomes the Compiler?, the scarce thing is shifting toward problem framing, domain understanding, taste, judgment, and knowing how to use AI as the execution layer. My claim is narrower than "university no longer matters". It is that for many founder-shaped kids, it may now be badly sequenced as the default opening move.
I should be upfront that this view is self-serving and shaped by survivorship bias. I did not go to university, so of course I see the build-first path more clearly than most people do. That does not make the argument false. It does mean it is easiest to execute for a narrow slice.
What I'd actually do
That also does not mean "skip university and sit in your room with ChatGPT". The failure mode here is drift. A bedroom, a few AI tools, and the identity of founder without the humiliation of selling, shipping, or being wrong is not the opportunity. It is avoidance with better branding.
If I were advising a founder-shaped school leaver, I would do something more concrete.
First, get them around serious adults and real work. If the family can make it work, a year overseas can help with perspective. Useful, not essential. What matters more is getting them out of inherited scripts and into environments where the consequences are real.
Then work inside a real business where the founder is still in the building. Not because corporate experience is useless, but because small businesses reveal the whole machine. Construction, logistics, agriculture, law, insurance, hospitality, aged care, trades, waste. The boring industries where the software is old, the workflows are ugly, and the money is real. I would care more about getting them around serious adults than about giving them a romantic founder identity.
Operator quality matters more than industry choice. A good operator in any of these sectors is the prize. A bad one is worse than no placement at all, because two years inside a chaotic or exploitative small business is not reps, it is wasted time and bad habits. The vetting work is the parent's, not the kid's, and it is not optional. Meet the operator yourself. Ask what they actually did with their last young hire, what the work will be, who your kid can go to when something feels off, and what would make both sides end the arrangement. Walk away if the answers are vague.
Once the placement is sound, the highest-leverage spend is simpler than most people think. Make sure your kid has access to a laptop that is not constantly getting in the way, and pay for the paid tiers of the frontier AI products they are actually going to learn on. The free versions are enough to play. They are not enough to build a serious working relationship with the models. I would want them learning more than one frontier product, not treating one as the winner. Part of the skill is understanding that the models are not identical, getting a feel for which one suits which task, and learning to cross-check work between them. The combination is the point: serious adults on one side, frontier tools on the other.
And when I say "learn how to think", I do not mean sitting around with abstract ideas. I mean learning a few durable concepts and testing them against reality: first principles, incentives, bottlenecks, trade-offs, compounding, reversibility. Shane Parrish's *The Great Mental Models* box set is a good place to start. So are Warren Buffett's shareholder letters and Jeff Bezos's shareholder letters. The point is not to collect clever frameworks. It is to build a better way of seeing what is actually happening, then use AI to turn that judgment into action faster.
Then pick a domain problem that actually bothers you and build around it with AI. Ship something ugly for one real user. Then another. Learn where the model helps, where it lies, where domain judgment matters, where trust matters, and where the work is still human. In a trade business, that might mean shortening quote turnaround, turning site notes into cleaner invoices, chasing debtors properly, or spotting which jobs are bleeding margin. AI is not doing the physical work. It is giving one attentive person much more operating leverage around the edges.
The test is simple. Are you getting closer to real work, real people, and real money, or are you drifting? If closer, keep going. If drifting, change the plan. By month twelve I would want five things on the board: a serious adult who would vouch for them, evidence they have handled real customers or real operational mess, something shipped that saved time or made money, at least one hard result, even if it is small, and a working arrangement that still looks healthy rather than exploitative. If none of that exists after a year, I would call it drift and push them toward a job or a degree.
This scorecard is also how you tell whether your kid is actually in the narrow slice. You cannot predict that at eighteen with any confidence. You can verify it by month twelve. If they are hitting the five, they were wired for this. If they are not, pivot early. That is precisely what the two-way door is for.
If after two years there is a specific degree, profession, or body of knowledge they genuinely need, go then. That is a completely different university decision from the one they would have made at eighteen, because now they have a real question the degree would answer.
What this is not saying
Some boundaries are obvious. If you want medicine, dentistry, nursing, law, engineering, teaching, or any profession where the degree is the actual licence to practise, go to university.
Some founder-shaped kids should also still go first. If the company you want to build depends on deep technical or scientific depth, a specific research environment, or a peer network that only exists inside a university, university first may be exactly right.
And many eighteen-year-olds are not ready for maximum freedom. Plenty would drift. Structure is useful when the person needs it. This post is not general career advice. It is a sequencing argument for the kid who is already showing unusual agency, can tolerate embarrassment, wants to be around real customers, and is deciding when to buy reps.
The question is not "is university good or bad?" The question is "what is the highest-value way to spend the next few years if founding is seriously on the table?"
The decision
My answer is that for a narrow slice of kids, university may now be better as move two than move one.
AI gives one person leverage that barely existed before. The broader market still has not fully reorganised around it. And for kids who are clearly wired this way, the switching costs of buying real reps first are often lower now than they are likely to be later.
If I were weighing this for my own kid, I would want them close to real problems, real operators, and the new tools, with a hard scorecard and a clean walk-back if the evidence does not appear.
This is my Three-Year Window argument at household scale. The macro opportunity for countries is made up of thousands of micro decisions like this one. If you are the parent of someone clearly wired this way, I think the most valuable thing you may be able to buy right now is reps before credentials, not instead of them.
I'm Ben Lynch. I write about founders, AI, and what happens next from New Zealand. Say hello at ben@thinkdorepeat.ai.
If you're new here, Start Here is the best place to begin.
If you're weighing this with a school leaver, send it to the one parent or operator whose judgment you'd actually trust in that conversation.