She has learned to talk to the model in a way that works. Whether this is a skill or a workaround is not yet clear.
Literal meaning: A prompt engineer is a person who specialises in formulating inputs to AI systems — particularly large language models — to produce desired outputs reliably and at quality. The skill involves understanding how AI systems interpret instructions, what kinds of formulations produce what kinds of results, and how to structure complex tasks for AI processing.
Origin: The term emerged rapidly from 2022–2023, as GPT-3, ChatGPT, and similar models became widely accessible and organisations began to recognise that getting reliable, high-quality outputs from these systems required specific skill. Job postings for “prompt engineers” with salaries up to $335,000 appeared in 2023, attracting media coverage. The term and the profession are already under pressure: as AI systems improve their ability to interpret natural language instructions, the specific technical skill of prompt engineering may diminish in value.
A profession built on the gap between what AI systems can do and what users can ask them to do — whose long-term existence depends on that gap staying open.
The Appeal: Prompt engineering captures a real skill. Understanding how AI systems process instructions, what failure modes to anticipate, and how to structure multi-step tasks for reliable AI completion — these require knowledge and practice. For organisations deploying AI at scale, the skill difference between a good and bad prompt engineer is measurable in output quality.
The Friction: The profession is temporally bounded by design. As AI systems improve, their ability to interpret poorly structured or ambiguous inputs improves — reducing the value of expert prompt formulation. That was measured early: Zhou et al. (2023) had a model write its own instructions and match or beat human prompts on 19 of 24 tasks. AI-Ninja is the person who has made this skill into an identity. When the skill goes, the identity goes with it. AI Dependency — structural reliance without critical reflection — is the consequence: organisations that have invested in prompt engineering as a core skill are dependent on the specific capability profiles of the AI systems they have designed for. Vibe Coder is the extreme case: outsourcing the translation entirely, accepting output whose logic cannot be interrogated. Cognitive Offloading becomes Deskilling when the expertise is in operating the AI rather than in the underlying domain the AI is operating in.
Why This Matters: Prompt engineer names the temporary value of a translation skill that exists because AI systems are not yet transparent enough to require no translation. The profession’s lifespan is a measure of how fast that transparency improves.
Related terms: Vibe Coder · AI Dependency · Cognitive Offloading · Deskilling · AI Literacy · AI as a Utility · Red Teamer · AI-Ninja
Read more:
- A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT — White, J. et al. (2023). arXiv
- The Prompt Report: A Systematic Survey of Prompt Engineering Techniques — Schulhoff, S. et al. (2024). arXiv