You Can Build Anything Now. That’s the Problem.

Everyone says you can build anything now. Technically true.

We skip the part about money (tokens).

The people who are actually good at using AI are the people who have tokens to waste. That's the thing nobody says out loud (if you follow the openclaw & claude-code guy; you know). Being good at this isn't a talent. It comes from running the same problem 10 different ways, watching 8 of them fail, and slowly learning what this thing responds to. You find your rhythm by burning tokens on experiments that go nowhere.

$20 a month doesn't buy that. $20 buys you enough to finish the work in front of you (well that's if your 5 hour rate limit doesn't ask you to pause). It doesn't buy the failed runs, the rebuilt prompts, the 5 hour session where you learn how far you can push an agent before it falls apart. You spend your budget on output, so you never get the practice.

This part used to be free.

Learning to program took the same thing: repetition, failure, hours of sitting with something broken until it made sense. But you paid for it with time. Your own hours. And everyone got the same allocation; a student in Bamenda and a senior engineer in SF both had 24 of them. If you were willing to spend your nights on it, you could catch up to anyone. That was the whole deal. Time was the price (bad back if you sit long enough)

Now the price is tokens, and tokens you buy. The practice itself has a bill attached. You can want it as badly as anyone and still not afford the reps.

So someone with more tokens isn't just producing more. They're learning faster. Every month they get better at this and you stay where you are.

You open x today; there's Loops. MCPs. Subagents. Some new workflow everyone swears by. And it all feels slightly off, because advice is worthless until you've tested it yourself. Until you have a variation or mixage of a few that will work for you.

Then the cost moves. A new model, a new agent, a new way of doing things, and what worked on $20 wants $100, $200, more.

If you can afford it, it's exciting. If you can't, it's scary.

Then the other half of the conversation: 'You can build anything now.' keeps popping; it simultaneous pulls you in the other direction; You can build anything but why am i not building anything; that messes with you (well i know it does to me) am i not good enough or am i doing this wrong.

but you pause and think; Can everyone build at the same level?

The labs are building everything themselves. Models, infrastructure, distribution, data, and the context users already handed them. Don't forget chips too. They ship in categories that belonged to SaaS companies and indie hackers.

Which puts indies & students in a strange spot.

I showed someone something I'd been working on (rondpoint.ai (s'ouvre dans un nouvel onglet)). She said it was good, it needed polishing. Then she said the big labs would make something similar soon, and she'd wait for that, since they already had her context.

That stuck with me.

How does an undergraduate in Bamenda who can't spare an extra $20 a month compete with a company spending millions on models, infrastructure, research, and tokens?

We keep saying you can build anything now. The better question: who can afford to keep building?

Good software still sells. But without room to experiment, I'm one more product in a category with a million of them. AI opened that door. Everyone can build, so everyone is building.

I hear the advice. Double down on a skill. Go deep. Position yourself so AI makes you faster.

I agree. But it's hard to go deep when the ground keeps moving. You can spend years getting good at design, research, writing, engineering, then watch part of it collapse into a prompt.

The skill doesn't become worthless. The bar moves & it costs money to practice. Building got cheaper. Competing didn't.

Here's what I think I'm landing on.

Building as we know it is probably finished. Not the act of it; the identity. The thing where knowing how to build was itself the value, and the rest was details. That held for a long time & what replaces it, I don't know.

But what I do know is that the people writing the advice are answering a question from before all this. Learn the skill. Ship fast. Build in public. That was good advice at some point but i'm not so sure anymore.

Other things like networking is still a constant; you still need for know man for some office for sell something (around here anyways)