Impact, Agency, and Taste
Among people who have already passed a high technical bar, the remaining differentiator is leverage — impact-per-hour. The bottleneck is no longer raw ability; it's the capacity to find and execute work that multiplies output.
High-leverage work is often not what's assigned: improving tooling that 200 people use daily, identifying a research direction before it's obvious, designing a system correctly the first time, or digging into data no one has looked at. Even apparently rote versions of these (e.g., manually scanning 1000 data points for anomalies) are often done by the most senior people, because they find severe issues others miss.
Two components drive leverage:
- Agency — the initiative to try, and the relentlessness to succeed. See making-success-inevitable for the specific distinction between doing work and taking accountability for outcomes. Key practice: work backwards from the root goal, not the assigned task. You often have more permission to reframe the work than you think.
- Taste — the quality of your predictive models: what will work, what won't, what matters. See taste-as-predictive-models for how to find where your taste is strongest and how to improve it.
Without agency, you won't make things happen even when you're pointed at the right thing. Without taste, you'll work hard toward the wrong thing.
Both are learnable through practice with reality feedback. The most important takeaway: you have permission to exercise both — most people wait for someone to assign them high-leverage work rather than finding it themselves.