Same as Ever

Morgan Housel's Same as Ever starts from a deceptively simple observation. The future is nearly impossible to predict in its specifics. Path dependency, compounding randomness, second-order effects. Nobody saw the iPhone coming, or COVID, or whatever's next. But human nature? That barely moves. The things that drive our decisions, our fears, our desires are ancient and remarkably stable.

I want to apply that lens to building software and startups, because AI has genuinely changed a lot about how things get built. The speed, the tooling, the economics. I've spent most of this blog writing about those changes. But this post isn't about what's changing. It's about what isn't.

You Still Need to Solve a Real Problem

This sounds obvious. It is obvious. And yet it's the thing that gets forgotten first when a new technology makes building feel effortless.

When building is hard, the difficulty itself acts as a filter. You don't spend six months building something unless you've at least convinced yourself there's a problem worth solving. When building takes an afternoon, that filter disappears. You can spin up a product on a whim. And many people are.

The number of apps, tools, and SaaS products entering the world right now is staggering. Most of them will fail for the same reason most products have always failed: they don't solve a problem that enough people care about enough to change their behaviour. That was true in 2005. It's true now.

Knowing What to Build Is Still the Hard Part

I wrote about this recently — AI is exposing who actually understands their product. The argument is that when engineering speed stops being the constraint, the real bottleneck becomes visible: product thinking.

I lived this at Airtasker. I joined as VP of Engineering and fixed the team's ability to deliver. We shipped more, faster. And the metrics didn't move. Because the problem was never engineering throughput. It was knowing what to build.

AI amplifies this. You can now build ten things in the time it used to take to build one. But if you didn't know which one mattered before, building all ten doesn't help. It just makes the confusion more expensive.

Design Is the Craft of Trade-offs

I need to be careful with the word "design" because it carries a lot of baggage. I'm not talking about pixel-perfect mockups or whether your buttons have the right border radius. AI can handle visual polish now. I've argued before that your favourite framework doesn't matter anymore. But the design thinking behind how you structure an experience for a user? That matters more than ever.

I'm talking about what Don Norman meant in The Design of Everyday Things. How is this thing crafted to help you do the thing it's intended to help you do? What information architecture makes sense? What do you leave out? What compromises are you making, and are they the right ones?

That kind of design — the kind that makes something usable, useful, and maybe even delightful — is a human skill. It requires understanding people, context, and constraints. It requires taste about where to draw the line. AI makes the surface layer cheap. Which means the deeper layer is where all the value lives now.

Distribution Doesn't Get Easier

There's an old line about first-time founders being obsessed with product and second-time founders being obsessed with distribution. AI is producing a lot of first-time founders right now.

Think about what YouTube did for video. It made creation and publishing essentially free. Anyone can upload a video. But making a living on YouTube is brutally hard, because the challenge was never creating content. It was getting people to find it, watch it, and come back. AI is doing something similar for software. It's democratised creation. It has not democratised an audience.

You can build a beautiful product in a weekend. Getting anyone to notice it, try it, keep using it, and tell someone else about it? That's the same grind it always was. If anything, there's more noise to cut through.

Startups Are Still Made of People

Every technology shift tempts us into thinking the technology is the thing. It never is. Startups succeed or fail based on founder dynamics, clarity of vision, the ability to sell, and whether the team can hold together when things get hard.

None of that changes because you have better tools. A founder who can't close a sale still can't close a sale, even if their product was built in a weekend. A team with unresolved tension will still fracture. A vision that isn't clear enough to align people will still produce scattered output.

When everyone has access to the same powerful tools, the differentiator is the people using them.

Learn Faster, Not Just Build Faster

AI makes building faster. But the bottleneck in startups was rarely building. It was learning. Understanding your users, testing assumptions, updating your mental model of the problem.

That learning loop is still limited by human cognition. You can ship a feature in hours now, but you still need to watch someone use it, notice where they hesitate, talk to them about what they expected, and sit with the discomfort of being wrong. That process doesn't compress.

The startups that win have always been the ones that learn fastest. AI changes the speed of building. It doesn't change the speed of understanding.

Simplicity Is Still Your Best Weapon

Every new technology makes it cheaper to add features. Which means every new technology makes it more tempting to add them. And complexity is still the thing that kills products.

The discipline of saying no, keeping scope tight, resisting the "we could also..." impulse? Timeless. And arguably harder now than it's ever been, because the cost of adding one more thing has dropped to nearly zero. But the cost to the user hasn't. Every feature they don't need is one more thing making the product harder to understand.

The Foundation

I want to be clear about what I'm not saying. I'm not saying nothing has changed. AI has changed an enormous amount about how software gets built, and if you're not adapting, you're falling behind. I've spent most of this blog arguing exactly that.

But the changes are happening on top of something. And that something hasn't moved. Solve a real problem. Know what to build. Design for humans, not for yourself. Find your audience. Build a team that works. Learn fast. Stay simple.

These are the laws of gravity of building products. They were true before cloud computing, before mobile, before AI. They'll be true after whatever comes next. The companies that thrive through this shift won't be the ones chasing every new capability. They'll be the ones that used new capabilities in service of these old truths.

That's the thing about gravity. You can build a rocket, but you still have to account for it.