Yesterday I wrote about working with AI like an army of lieutenants—how we can now coordinate work at the scale of a small team. Today, let’s talk about the real constraint. 🧵

The bottleneck when operating alone isn’t the AI. It’s you. 👈

The AI can generate at 10x or 100x speed, but you still need to review everything. And we’re not trained to review at this scale. You become the limiting factor—the human in the loop who must evaluate, refine, and certify each piece of output.

You might generate a week’s worth of content in a day, but can you meaningfully review it all? Can you verify the logic in five different code modules? Can you ensure consistency across a dozen documents? Can you spot the subtle errors in rapid-fire output? 🔍

This is a genuine constraint. Until we develop automatic ways to validate certain types of output—or trusted frameworks that reduce the need for complete human review—your review capacity caps your actual throughput.

📜 Some historical context: The 9-to-5 corporate job is actually a post-industrial invention. Before that, most people ran small operations—working alone, with family, or with a few employees. Growing up in a developing country, I saw this firsthand. Even today, much of the world works this way.

The difference? Those operations were limited by human capacity. One person with a few helpers can only scale so far.

With AI tools, we can generate at unprecedented scale. But we’re discovering that review capacity is the new constraint. We can produce like a team, but we still evaluate like an individual. ⚖️

💡 What this means: We need to develop new skills beyond just prompting and orchestration:

• Rapid quality evaluation techniques • Knowing when to trust the output and when to dig deeper • Building systematic review frameworks • Creating validation checkpoints for different types of work • Learning to spot patterns in AI output that signal quality issues

These skills take time to build, especially when there are few established patterns to follow.

We’re at the beginning of figuring this out. The next breakthrough isn’t just better AI—it’s better ways for humans to work with AI at scale. 🚀