Impact, agency, and taste — Ben Kuhn

Among highly screened technical talent, the biggest remaining bottleneck is often the ability to get leverage — work with a big impact-per-hour multiplier.

Examples of high-leverage work:

Improving tooling, documentation, or dev loops (tiny fix saves hundreds of user-hours).
Identifying promising research directions (small bet that became major influence).
System design (small part of time but makes the final system much better).
Collecting and digging into data (often reveals huge opportunities; even "rote" versions are done obsessively by seniors).

Finding high-leverage work has two components:

Agency: initiative/proactiveness to make things happen + relentlessness and resourcefulness to ensure success.
Taste: good intuition for what will and won't work well (in the large: picking important problems; in the small: picking approaches that will work).

Without taste, you work toward the wrong thing. Without agency, you get nowhere on the right thing.

Improvement comes mostly from practicing with good feedback loops from reality. The most important takeaway is that you have permission to try to exercise good taste and high agency.

Specific agency advice: understand and work backwards from the root goal. Don't just follow the assigned task if a better way (or not doing it) serves the higher goal better (example: suggesting to postpone a scheduler project because of maintenance burden on an overloaded team).

See its place in the generalist reading list under Strategy: a-reading-list-for-generalists-dylan-bowman.

The framework is captured in impact-agency-and-taste.