Akshay Sinha

Akshay Sinha / Writing

AI sets the length. You set the direction.

Building got cheap. Judgment didn't. What still decides whether a company goes anywhere once AI does the building.

Oct 2026 · Operating

Start with one number: revenue per employee.

Large IT services firms earn tens of thousands of dollars per person. A handful of crypto-native teams, Hyperliquid the best known, have been reported at around $100 million in revenue per person, with headcounts in the low double digits. Same metric, orders of magnitude apart.

The difference is not effort. Companies that add AI to an existing structure get incremental gains. Companies built around it from the start get a different shape. Converting to AI and starting with it are different games.

A beta in a weekend

A beta is not just code. It is a spec, a design, a website, marketing and a sales deck. That used to take five people a quarter, and most of the quarter was coordination: meetings, waiting, handoffs, rework. Five people have ten lines of communication between them.

Now one person with AI can cover every function over a weekend. No handoffs, no coordination cost.

But the weekend only works if you already know what a correct beta looks like. AI compresses the execution, not the judgment. Without that experience, the same weekend produces something that looks finished and is wrong.

Three limits

Anyone can do it. If one person can build it in a weekend, so can everyone else. Whatever you ship can be copied in weeks. The question moves from "can you build it?" to "why would anyone pick yours?"

It is confidently wrong. AI fills gaps with assumptions and states them with confidence. It only works when you already know what the right outcome looks like.

It has no direction. Picture where you are trying to go as a line, and yourself as an arrow. Progress is how far you travel along the line. AI makes the arrow two or three times longer. It does not change where it points.

Same multiplier, different angle

Point the arrow at a customer you understand, and AI ships in a weekend what used to take a quarter. Point it slightly off, and you get a lot of output that goes nowhere: forty polished slides, features nobody asked for. Busy, not moving. Point it the wrong way, at a wrong assumption or the wrong customer, and AI takes you further away, faster, and it looks convincing the whole time.

Before AI, a wrong direction cost you a little. Now it costs you three times as much.

What stays scarce

Distribution. Getting a stranger to pay you is harder now that everyone can build.

Trust. Someone still signs the contract and owns the failure.

Judgment. The angle of the arrow.

If you already own the scarce part, whether that is customers, suppliers or credibility in a market, AI is the execution team you never had.

What to do with it

If you start something, assume building is free for everyone and win on distribution and domain knowledge. If you join a company, expect routine analyst work to go first; own outcomes and direct the AI.

The short version: building is cheap, judgment isn't. AI sets the length, you set the direction. It can pull your weak skills up and your strong ones higher, but only in the direction you chose.