Table of contents
Open Table of contents
Forget the 10x engineer
The industry spent a decade arguing about the mythical 10x engineer — the lone genius who out-produces ten peers. It was always a flawed idea: software is a team sport, and heroics don’t scale.
The AI era quietly retires that debate and replaces it with something more interesting: the 10x team. Not one superhuman, but a small group of senior people who each direct fleets of agents — and whose collective output looks like a much larger org used to.
The shape of the team changes. A handful of people who can specify well, review fast, and hold the system in their heads, each amplified by agents doing the implementation. Fewer hands on keyboards. More judgment per head.
Judgment is the scarce resource
When implementation capacity is effectively unlimited, the binding constraint becomes human judgment — the ability to decide what’s worth building and whether what got built is actually right.
This flips a lot of conventional org design. The instinct to “add more engineers to go faster” weakens, because raw implementation throughput isn’t the bottleneck. What you need more of is people who can direct throughput well. A team that’s all juniors and agents produces a lot of confidently-wrong code, fast. A team with a few strong seniors and agents produces leverage.
A team that’s all hands and no judgment just generates plausible mistakes at scale.
What changes for engineering leaders
If you run an engineering org, a few things shift under your feet:
- Hiring bar over headcount. The marginal value of one more “can write code” hire drops. The value of one more “can decide and verify” hire rises. Smaller, more senior teams beat larger, junior-heavy ones.
- The juniors question is real. If agents do the grunt work juniors used to learn on, how do people grow into seniors? This is the hard unsolved problem. The answer is probably deliberate apprenticeship — juniors learning to direct and verify early, not just to type.
- Context becomes infrastructure. A 10x team running on no shared organisational brain is just a fast way to fragment. The knowledge layer is now a leadership responsibility, not a nice-to-have.
- Throughput stops being the metric. Measuring “story points shipped” in a world of unlimited implementation is measuring the wrong thing. Outcomes, quality, and judgment-per-decision matter more.
The team as the unit of leverage
The through-line across everything I write about AI engineering is that leverage now lives in the team, not the individual. The orchestrator mindset, the discipline of shipping verifiable outcomes, the organisational brain — they all compose into a small group that punches far above its headcount.
That’s the 10x team. Not a person to hire, but a way to organize. And it’s the bet behind my whole AI Engineering practice.