prompt clarity

Before/after examples show direct correlation between prompt clarity (outcome-based vs process-based) and agent behavior.

Natural language constraint ('use gh CLI') outperforms formal tool definitions, suggesting clarity/simplicity in prompting matters more than comprehensiveness.

The 'knowing what you want' principle is the highest-level form of prompt clarity—intent must precede wording.

Agent-based architecture forces clarity about agent responsibilities. Each agent needs a clear role and scope, which is a form of context clarity.

Author emphasizes that describing what you want in plain English enables professional outcomes. Suggests clarity of problem specification directly enables context quality.

Harford's framing implies LLM plausibility is an inherent trait, but practitioner experience shows that clear problem definition and proper context dramatically improves output quality. The article do

Author's frustration likely stems from unclear problem specification ('here's a task, make code'), but they don't recognize this as a lever for improvement. The complaint suggests they're not applying

Article implies that clear thinking (and thus clear problem articulation) is better than AI-generated outputs. Weakly supports thesis that clarity is bottleneck, but doesn't advance HOW to achieve it.