The 'Prompt' Is Not a Skill — And We Need to Stop Pretending

This post makes the case that prompt writing has been dressed up as a discipline when it is functionally just task description — you say what you want, the model figures out the rest. The provocation is not new; the prompt-engineering-is-real vs. prompt-engineering-is-hype debate has been running since GPT-3. What the post does usefully is point toward where the harder, more durable skills actually live: evals, context management, model selection, and knowing when to not use an LLM at all. For a founder deciding whether to hire a prompt engineer or an AI engineer, that framing is more useful than the clickbait title suggests. The writing is punchy but the argument would benefit from concrete examples rather than assertion. Read it as a short reset if your team has started treating prompt templates as precious proprietary IP. -> Best for: solo founder or technical PM hiring their first AI-adjacent role