In the past few weeks, I’ve spoken with several founders about how much AI coding tools are creating enormous leverage for the people who actually take the time to master them. We need new frameworks for development. Waterfall was used for the mainframe and PC era and gave way to agile in the internet and cloud era. Similarly, we need a different development framework in the age of AI.

At a recent board meeting for Company A, led by an AI-native CTO we recruited away from Company M, the data was stark. The top 5% to 10% of builders are now 3 to 5 times more productive than they were a year ago. The median builder is up maybe 10% to 20%. We are seeing the same pattern across multiple companies. Naturally, founders want to understand what separates the two groups.

To better understand the changes in software development, I asked a founder to show me something she built over the weekend using Claude Code. This was not a prototype. It was a fully functioning product with a backend, a polished interface, and integration into the company’s database. Two years ago, it would have taken a small team weeks just to get a rough version into shape. Yet, she is not a software engineer. She simply saw a problem no one was solving and decided to build the solution herself.

What stood out was how she described the process. She was not coding. She set intent, reviewed output, iterated, and refined until the output matched her vision. Many developers take pride in the code they write. Builders take pride in what gets built. She did not care about the code. She took pride in commanding a fleet of agents to realize her vision.

It reminded me of Ender’s Game, one of my favorite sci-fi novels. Ender Wiggin’s arc is instructive. At Battle School, he stops trying to be the best individual pilot and instead reorganizes the Dragon Army into autonomous toons, small units empowered to make decisions without waiting for orders. His army becomes dominant not because he does more, but because he delegates better.

At Command School, he never flies a single ship. He sits at a simulator directing fleet commanders who direct their own squadrons. His job is to make strategy, exercise judgment, and make hard decisions in the face of uncertainty. This maps directly to what is happening in software.

The best builders today do not try to write every line of code. They use their taste and judgment to determine what to build and plan how to build it. Then they spin up parallel agents: one refactoring, one writing tests, one working on the UI, all operating with clear intent. Cursor can run 10 to 20 cloud agents simultaneously, each producing a pull request. Claude Code orchestrates sub-agents across entire codebases. The developer who insists on controlling every line of code is Bonzo Madrid, rigid, hands-on, and ultimately outmatched by someone who knows how to delegate with trust.

In Ender’s final battle, outnumbered and overwhelmed, he wins not through better tactics but through imagination, sacrificing his entire fleet in a bold, unconventional strike at the enemy’s home world. It was a decision born of conviction, not incremental optimization. For now, in software, that is the product vision. AI can generate the code. It cannot yet imagine the product that doesn’t yet exist in the world.

Finally, Ender believed he was in a simulator. In reality, he was commanding the actual live fleet. We are in a similar moment. Claude Code, Codex, and Cursor are not toys, sandboxes, or simulations. They are the real interface through which production software is being shipped today. Treating them as anything less misses the point.

Ender did not win because he was the best pilot. He won because he saw patterns others missed, made decisions others would not, and trusted his team leaders to execute while he focused on what only he could do. The real question is: what kind of fleet commander will you be?