The falling cost of intelligence

There is a curious paradox brewing in tech right now: as intelligence becomes more artificial, it becomes more affordable. Last week OpenAI slashed the cost of its o3 model by 80 percent, from ten dollars to two per million tokens.  No trade-offs. Just better, cheaper thinking.

The Falling Price of Thinking

This is shift in the economics of cognition. You hired the smartest people you could find, paid them well, and hoped they'd solve your hardest problems. Intelligence had natural limits, biological ones. A brilliant engineer could only think so fast, work so long, process so much information. The best minds commanded premium prices because there were only so many of them to go around.

But what happens when intelligence becomes abundant?

The cost curves now look like storage in the early 2000s: down, down, and down again. Today you can rent reasoning that once required a team of PhDs for the price of a latte. Tomorrow it will be cheaper still. Intelligence is drifting toward the same commodity shelf as bandwidth and electricity.

From Scarcity to Surplus

When a resource flips from scarce to abundant, value migrates. The moment electricity became cheap, fortunes shifted to appliance makers. When servers got cheap, software ate the world. Cheap intelligence pushes value higher up the stack toward judgment and increasingly, toward taste.

The New Scarcity: Questions

This creates a strange inversion. As intelligence gets cheaper, the premium shifts to something else entirely: knowing what questions to ask.

The real bottleneck isn't processing power anymore, it's problem identification. In a world where answers cost pennies, the ability to frame the right question becomes priceless. The companies and individuals who'll thrive aren't necessarily those with the smartest algorithms, but those with the sharpest instincts about what problems are worth solving.

Think about what this means for expertise. The traditional knowledge worker's value proposition, knowing stuff, is rapidly eroding. But a new kind of expertise is emerging: the ability to choreograph intelligence, to know which tools to use when, to orchestrate multiple AI systems toward fresh outcomes. It's less about being smart and more about being wise about intelligence.

The Final Mile: Taste

Yet there is another layer that algorithms still fumble: the subtle, human sense of what feels right. Call it taste, call it editorial instinct, call it the last two percent. Whatever the label, it is becoming the key differentiator when everyone has access to the same AI firepower.

What Taste Looks Like in Practice

• Images – Type “vintage travel poster, sunset over mountains” into Midjourney and you will receive dozens of passable options. Only a few will capture the warmth and grain that make the style sing. Choosing the right starting point and iterating over it becomes the new craft.
• Landing pages – Generative site builders, and component libraries spit out layouts that check every UX box. The pages that convert, though, sweat the spacing of buttons, the cadence of headlines, the nuance in the call‑to‑action. Taste is the difference between a template and a brand.
• Micro‑apps – Low‑code platforms let anyone stitch APIs together. The best apps hide complexity, surface sensible defaults, and tell a story through their flow. Taste often shows up in what you leave out.

Democratization and the Anxiety of Sameness

A startup in Lagos can now access the same reasoning muscle as Goldman Sachs. If everyone has the same grade of AI, how do you stay ahead? The answer is to climb the stack: away from raw output, toward creative composition, toward unexpected combinations, toward taste.

This shift echoes every major commoditization wave we've seen. When electricity became abundant, the value moved from power generation to what you could actually do with that power. When computing got cheap, the magic wasn't in the processors but in the software. Now, as intelligence becomes commodity, the differentiation lies in taste, judgment, and our peculiar human ability to spot patterns that actually matter.

Where We Go Next

We are entering what might be called the post‑intelligence economy, not because thinking stops mattering, but because it stops being scarce. Intelligence is getting cheap. Judgment, creativity, and taste remain expensive.

Three questions to take home:

1. If answers are free, which questions matter most to your business?
2. What would your product look like if you rebuilt it with a tenth of the headcount but twice the taste?
3. Where can you replace frantic output with deliberate editing?

Intelligence is drifting toward free. Wisdom, meanwhile, grows more valuable by the day. The edge now is not in buying more answers, but in choosing better questions, and acting on them with taste.