dshan

Think it, cook it, learn, think again, cook again, learn.

essay // publishedmdx
/2 min readaistrategywriting

Good, Fast, Cheap

A short argument that AI has changed the old tradeoff between quality, speed, and cost.

For a long time, the rule felt fixed: pick two.

Good, fast, cheap was treated like a law of nature. If you wanted quality, you paid in time. If you wanted speed, you paid in money. If you wanted to keep it cheap, something had to give.

That made sense when effort scaled linearly. Better work meant more hours. Faster work meant more people. Cheaper work meant less polish.

Then the tools changed.

Today, modern LLM-based systems create a different kind of leverage. They do not just automate narrow tasks. They can draft, structure, synthesize, and iterate at a speed that would have been impossible a few years ago.

The tedious middle of many creative and analytical jobs is now much cheaper to produce.

That does not mean judgment disappears. It does not. Taste still matters. Good editors still matter. Domain expertise still matters. The difference is that the scaffolding can be generated quickly enough that the old constraint is no longer absolute.

The old constraint

The old model assumed human work was the bottleneck. To make something better took longer. To make it faster meant adding more expensive help. To keep the cost down, the system had to accept a compromise.

That was not a myth. It was the operating reality.

The new constraint

AI changes the base rate of iteration.

Drafts can appear much faster. Entire workflows can compress. But the result is only good if someone knows how to shape the output, refine it, and push it in the right direction.

That means the human part of the job shifts from producing every layer to guiding the system with good questions, strong taste, and context.

Earn all three

So yes, good, fast, and cheap is now possible.

But not automatically.

You do not get it by default. You get it by knowing how to use the tools well enough to keep the output sharp.

The old Venn diagram is broken.

The new version is earned.