Prompt engineering as a standalone job title was largely a temporary phenomenon, and those headline-salary listings have mostly disappeared. But the underlying skill is real and has quietly been absorbed into other roles — which is the usual fate of a genuinely useful skill that isn't a whole profession.
Why the standalone role faded:
1. Models got substantially better at understanding ordinary instructions, so the elaborate incantations that used to matter now matter much less.
2. It turned out to be a skill, not a job — like 'good at spreadsheets' or 'good at search'. Valuable, learnable in weeks, not a career on its own.
3. Companies discovered the hard part wasn't writing prompts. It was everything around them: evaluation, data pipelines, retrieval, error handling, cost control, integration. Those need engineers.
What replaced it, and where the money actually is: **AI engineer** or **AI application developer**. That role does include prompt design, but mostly it's ordinary software engineering aimed at LLM systems — building retrieval over company documents, designing tool use, writing evaluation harnesses to measure whether output quality regressed, managing latency and cost, handling failure modes. Strong demand, genuine depth, and it's reachable from a normal software background.
The skills that are durably valuable:
- **Evaluation.** Knowing whether a system's output is actually good, systematically rather than by vibes. This is the scarcest skill in the space and the least glamorous.
- **Decomposing a fuzzy task into steps a model can do reliably**, with checks between them.
- **Knowing what these systems can't do**, so you don't build a product on a capability that doesn't exist.
- **Grounding techniques** — retrieval, tool use, structured output — that reduce hallucination.
- **Domain expertise.** Applying this well in law, medicine, education or finance requires knowing the domain deeply. This is where the least replaceable people are.
So the practical advice: don't pursue 'prompt engineer' as a title. Become a software engineer who builds AI-powered products, or a domain expert who deploys these tools expertly in a real field. Both are durable. The prompt-writing skill itself is worth a couple of weekends of deliberate practice, and then it's just part of how you work — which is exactly what happens to most tools that turn out to matter.