You Go Conquer the World — I'll Build What Comes After

You go conquer the world — I'll build what comes after.

Using AI these days stirs an odd feeling in me.

It's good.

Really good.

Set it to development work and it's no longer just spitting out a few lines of code.

It understands structure, edits code, catches bugs, sustains long tasks, and genuinely helps build real products.

Sometimes it even feels far faster than working with an actual person.

But there's one problem.

Ask Claude for Korean copy and my blood pressure spikes.

The same guy who was coding so brilliantly suddenly

produces sentences like “Discover new possibilities through an innovative experience.”

It just spits that out.

In that moment I think:

'Ah, this punk is at it again.'

The funny part is that I can't get rid of him.

He's too good at the job.

Mid-development, I find myself in awe.

“Wow, he's genuinely good.”

Then I ask for one line of product copy and the cursing starts again.

“Why on earth do you write Korean like that?”

I put him back on development.

Brilliant again.

The relationship ends up strange.

He's like a key employee who does excellent work and gets chewed out constantly.

There are many moments I want to fire him, and absolutely no way I ever could.

So the conclusion is simple.

He keeps the development work.

Final copy approval is revoked.

But think about it — copy is by no means a small matter.

Copy is not the product's decoration.

It's the interface.

Users never know how beautiful the internal code is.

What users see, in the end, is a few lines of text.

What this service is.

Why they should use it.

What to press right now.

Whether it's safe to pay.

However good the features, if the sentences are off, the whole product looks cheap.

Conversely, attach one precise sentence to the same feature and the product's value snaps into focus.

So for AI development tools to be truly complete, coding well isn't enough.

They have to follow the entire way actual people work.

And what I want from AI companies these days isn't complicated.

Before any grand AGI declarations, I'd like this first:

a model that codes better than Claude in real work,

costs less than Kimi,

writes natural Korean,

doesn't lose context on long tasks,

and spouts less nonsense when debugging.

Build that. That's the request.

Honestly, from a developer's standpoint, who the world's best AI company is doesn't matter much.

Who declares the grander vision doesn't matter either.

In the end there are only a few questions.

Does it do what I asked, properly?

Does it keep its bearings in a large codebase?

Does it hold context when given hours of work?

When something breaks, does it find the actual cause?

And how much does it cost?

When performance and price are overwhelming, people don't hold long philosophical debates.

They just pay.

So what I want to say to OpenAI is simple too.

Altman — wake up.

Ship something that develops better than Claude and costs less than Kimi.

And if you pull that off?

World conquest?

Yeah.

Go ahead.

Do as you please.

I'm not particularly interested.

I'd just like you to hurry.

Because I have the next generation to build.

If the AI companies build better models,

drive down computing costs,

raise coding ability,

and make it possible to hand over longer stretches of work,

then I take that and build new systems on top of it.

When tools advance a generation,

the level of product you can build with them rises a generation too.

This no longer feels abstract.

I can see it.

I can grasp it.

If the model gets a little better today, what I can build today changes.

Work that took days a few months ago becomes hours,

and work that needed several people a few years ago becomes something one person can do.

So if AI-talk used to be closer to futurology,

today's AI is closer to a means of production.

Not “someday, such a world will come.”

It's not that.

It's “wait —

this works right now?”

That's what it's closer to.

That difference is probably the biggest one.

When the future is far away, you can wait.

But when the future starts appearing before your eyes, waiting becomes hard.

Because you can see what you could build if it got just a little better.

So I keep prodding the model companies.

Build faster.

Make it cheaper.

Make it better.

Cut the useless nonsense.

Get Korean right, too.

Whether you conquer the world or not — handle that yourselves.

I'll take the tools and build the next thing.

Maybe the most interesting people in today's AI race aren't only the ones building the models.

The people who take those models

and build something new on top of them.

I find that side far more fun.

So.

Sam Altman.

You go conquer the world.

I'll build the generation that comes after.

Just hurry it up.

#AI #ArtificialIntelligence #GenerativeAI #AIDevelopment #AICoding #Developer #SoftwareDevelopment #Claude #Kimi #OpenAI #ChatGPT #SamAltman #AIAgents #DevAutomation #Productivity #Startup #ProductDevelopment #FutureTech #AIEra #TechFuture