Prompting just split into 4 different skills. You’re probably practicing 1 of them (+ 7 prompts and a pre-flight to close the gap)

Source: Nate’s Newsletter Date: 2026-02-27 URL: https://natesnewsletter.substack.com/p/prompting-just-split-into-4-different

Summary

Nate argues that as models shifted from conversation partners to long-running autonomous workers, “prompting” diverged into distinct practices that require different skills: specification engineering, intent framing, eval harness design, and constraint architecture. Most practitioners are still doing chat-based prompting — iterating in real time to get 70% results — while the gap to fully autonomous, finished-work output requires upfront specification and context engineering rather than conversational refinement. The piece offers a pre-flight diagnostic and seven concrete prompt templates illustrating the difference.

Implications

• The chat-to-specification transition is the practical skill boundary between “AI-assisted” and “AI-automated” workflows — relevant for evaluating where current tooling sits on that spectrum.
• Eval harness design appearing as a first-class prompting skill signals that quality assurance is moving upstream into prompt authorship, not downstream into human review.
• The Klarna failure case (good context, unclear intent) is a useful calibration: specification completeness matters more than specification length.
• Feeds the thread on workflow structure as the binding constraint: context engineering is where agent reliability is won or lost before the model is ever invoked.