Beyond Instructions: GPT-5’s An AI Upgrade You Didn’t Ask For.

GPT-5

I’d call it a Mixed-Fruit Jam.

Because right now, OpenAI is losing ~$2.25 for every $1 they earn.

Unit economics?

Broken.

So they’ve reached for an old playbook - the Free Candy strategy:

Hand out incredible features for cheap or free, make you dependent.

Slowly fence it off.

Sell the illusion of progress… while quietly optimizing for cost.

So what’s different about GPT-5?

It’s a router.

You type a request, and somewhere deep in the server racks, a quiet little algorithm decides:

“This is easy - give them the cheapest model possible.”
“This one’s tough - okay, spin up the expensive one.”

Efficient - but a double-edged optimization.

Most people will never know what really changed this week.

Even fewer will understand why it matters.

But once you see it, you can’t unsee it.

Because GPT5 isn’t just waiting for instructions anymore.

It moves first - without letting you tune it.

It decides when to think… and when not to.

It fills in gaps you never asked it to.

That’s thrilling… until you realize:

When AI starts suggesting goals, not just executing them -

whose values are baked into those goals?

For most casual users?

They’ve just been vaulted one and half year into the future - without even knowing “model switching” existed.

This is an upgrade.

For power users - developers, researchers, creatives - it’s like buying a Swiss Army knife and being told:

“Good news! We welded it into a single blade so you never have to choose again.”

You didn’t just lose a tool.

You lost that bro who let you drive.

So we’re left with a few truths:

• If you want loyalty → give people power, then make them want to give it back.

• If you want mass adoption → hide complexity until the awe sets in.

• If you want to survive → optimize margins before the venture money runs out.

This isn’t the iPhone 1 moment.

It’s the iPhone 16 moment.

And that leaves us with the oldest tension in tech:

Do you build for love, or for scale?

Serve the devoted, or serve the masses?

What would you choose if you were OpenAI - love or scale?

Originally shared on LinkedIn