In early 2023, a handful of AI labs started posting for a role almost nobody had heard of: prompt engineer. MIT Technology Review later profiled the title as one that arrived with six-figure salaries and then grew into something bigger — solutions work, retrieval design, system-prompt architecture. Two years on, the title has split, renamed itself, and in some companies been absorbed into a broader job called AI engineer.
That churn is the story of computing hiring right now. If you're trying to break in, the job you're aiming at may not have existed when you started studying for it, and the one you were told to aim for may already be folding into something else.
The Titles Are Moving Faster Than the Curriculum
Look at what's actually hiring. LinkedIn's 2026 Jobs on the Rise report puts AI engineer — sometimes labeled machine learning engineer — as the fastest-growing role overall, with common skills clustered around LangChain, retrieval-augmented generation, and PyTorch. The same list surfaces AI consultants, AI researchers, and data annotators near the top.
None of those titles map neatly to a college major. Universities are still catching up. Bootcamps rewrite curricula every few months, and employers keep splitting one job into three and then merging two of them back together.
If you want a fuller sense of where the demand actually is in tech careers sits in tech careers, the older reference lists still hold up on the fundamentals — networks, databases, security, systems — even as the labels on top keep shifting.
Chasing the Newest Title Is the Obvious Move — and Usually the Wrong One
The instinct is to sprint at whichever job posting looks hottest this quarter. Prompt engineer was that job in 2023. Forward-deployed engineer became that job in 2025. AI agent architect is having its moment now.
The problem is timing. By the time a title trends, the market is already reshaping it. Prompt engineering, as a standalone role, largely dissolved into other jobs within eighteen months, and the skill outlasted the label. Candidates who trained narrowly for the title found themselves losing out to generalists who could do the prompt work AND ship the surrounding system.
There's a second trap. Employer job descriptions in fast-splitting fields are often written by someone who isn't sure what they want yet. "AI engineer" at one company means model fine-tuning; at the next it means API integration; at a third it means client-facing solutions work. Studying for the title is studying for a moving target.
Aim at the Work, Not the Label
The move that works is boring and durable: build skill in the underlying work, then translate that skill into whatever title the market is using this quarter. A few practical anchors:
- Pick the layer, not the buzzword. Decide whether you want to build models, deploy them, integrate them into products, or evaluate them. Those four layers stay stable even as titles churn on top of them.
- Ship something public. A working repo, a written case study, a small deployed app. Employers hiring into fuzzy new roles lean on portfolios far more than credentials, because no credential has caught up yet.
- Learn the adjacent boring skills. Version control, testing, cloud basics, SQL. Every newly-invented AI role still runs on this substrate, and candidates who skip it get stuck at prototypes.
- Read job posts as evidence, not instructions. If ten "AI engineer" listings describe ten different jobs, the role is unsettled. Apply where the described work matches what you actually want to do.
The Non-Traditional Path Is Now the Common One
Candidates landing these newly-minted jobs frequently don't have a CS degree in the specific area, because the area is too new for a degree to exist. They come from adjacent software work, from data analysis, from product, sometimes from operations or research in another field. What they share is evidence: a portfolio the hiring manager can click through, contributions to an open-source project, a blog post that walks through a real system they built.
That path is more open than the job board makes it look. Plenty of those openings are being filled by people whose résumés don't mention the role by name, because two years ago, the role didn't have a name.
How to Make a Career Decision This Year
The safest bet is the one where the underlying work compounds, not the one with the safest-sounding title. A computing career built around models, systems, and evidence of shipped work will survive three more rounds of renaming. A career built around whichever title trended in the quarter you enrolled in a bootcamp probably won't.
Watch the titles. Read them for what they say about where the market is heading. Then aim a step deeper than the label.
