What You’ll Learn

Here’s a puzzle I keep chewing on: AI tools keep getting better and cheaper, yet everyone’s output is starting to look the same. A video I came across (from another creator, breaking down an engineer’s real experiment) nailed why — and it reframed how I think about where educators should spend their AI effort. This tutorial is my take on that lesson, aimed at people building a teaching business.

By the end you’ll understand why cheap and expensive models now tie on ordinary work, where the real value moved, and a simple two-layer way to organize your own AI use so your work stops looking like everyone else’s.

The Setup: Cheap Model Ties Expensive Model

Mitchell Hashimoto — a deeply respected engineer — tested Fable 5, the pricey frontier model, against much cheaper models on ordinary work: “implement this feature,” “build this thing.” All three produced equally acceptable output. The budget model cost under a dollar and finished in minutes; Fable 5 cost 9x more for the same result. That launched a thousand takes: route everything to cheap models.

And you should route a lot of execution to cheap models. But here’s the part almost nobody said out loud.

“When everyone in AI is saying the same thing at the same time, that’s exactly when to ask: where does value move once we’ve all figured this out?”

The Twist: The $40 Question Nobody Assigned

Hashimoto ran one more test. He handed the frontier model a problem the cheap ones couldn’t touch — optimizing a gnarly piece of systems code he’d written himself. Two hours, $40, and it reached a level of performance he says he couldn’t have hit on his own.

The question that matters: who assigned that task? It wasn’t on a backlog. No manager prioritized it. It didn’t exist until an expert suspected something new had become possible and spent money to find out.

💡 In Plain English: AI can only do work someone has imagined. The cheap tools execute; they don’t decide what’s worth executing. So the ceiling on what AI is worth to you was never the model or the price — it’s the size of your list of things you know how to ask for.

Why Everything Looks the Same

“Implement this feature” is work everyone already knows how to ask for — and that’s exactly where the models have converged. Now zoom out from models to people: we share prompts, follow the same channels, run the same playbooks. When a million of us run the same tasks through the same tools, of course the results converge. AI didn’t make your work generic; it revealed that differentiation is a human task.

It’s the BlackBerry story. BlackBerry executed brilliantly — best keyboard, best email — inside a category everyone had already imagined. Apple imagined a different answer to what a phone even was. Same industry, comparable execution muscle; imagination set the multiplier, and one company’s execution became worth a hundred times the other’s.

The Test: Has Your Task List Changed?

Here’s the gut-check, and it works for a person as well as a company. Has what you ask AI to do changed in the last 12 months? The last six? The last three? Or are you running your old list faster and cheaper and calling that transformation?

If it’s the old list, nothing’s wrong with your tools — you have an imagination shortage, and you’re pouring money into optimizing execution in a commoditized market.

The Good News: Imagination Isn’t a Gift, It’s Fingertip Awareness

“Imagination” sounds like something artists have and the rest of us don’t. That’s not what was operating in the $40 story. Hashimoto could pose that question because he has thousands of hours inside these models — he knows where the capability line has moved by instinct, not from a benchmark chart. You can’t imagine a use for a tool you’ve only read a summary of.

This is where most of us quietly sabotage ourselves: we point AI at work we already have and ask if it can go faster or cheaper. That’s pointing the telescope at the ground. The better question is: what can this do that I could never even ask for before?

The Two-Layer Stack — For Your Teaching Business

The practice is almost embarrassingly simple, and it maps to two layers:

Execution layer (cheap models): your daily grind — repurposing content, drafting emails, summarizing, formatting lessons. Optimize this aggressively and drive the cost down. (This is exactly the local-model routing I use for transcript work.)

Imagination layer (frontier models): the surgical questions that change what the execution layer is even building. This is where your scouting hours go. And these aren’t artsy tasks — they’re technical and business imagination: looking at a teaching problem in a new way because you have fingertip awareness of what a frontier model can now do.

The two aren’t in competition. Cheap execution is the engine; frontier imagination is where you steer. And the cheaper execution gets, the more valuable every frontier question becomes.

Don’t Bolt It Onto the Old Layout

When factories electrified, the tech worked on day one — but the payoff took decades, because they kept the steam-era layout and just bolted a motor where the engine used to be. The gain came when someone redesigned the whole building around what cheap distributed motors made possible. AI is the same: most people are running their existing task list through cheaper models and reporting the savings. The savings are real — and available to every competitor. That’s table stakes, not leverage.

The Takeaway (My Angle for Educators)

Route your daily teaching-business execution to cheap models — absolutely. But then take your scouting hours seriously: spend frontier-model time on the questions no one in your niche is asking yet. Ask “what can I now do for my learners that was impossible six months ago?” That’s where your work stops looking like everyone else’s — and in the companion tutorial, I take this straight at a concrete idea for education.

Teach more, let the agents do the rest — but do the imagining yourself.

This tutorial is my perspective on ideas from another creator’s video; credit to the original source.