We've Heard This Before
Every few years, Artificial Intelligence has a moment. It arrives with breathless press coverage, inflated expectations, and a wave of investment. Then reality sets in, the hype deflates, and we enter what researchers call the "AI winter."
It happened in the 1970s. Again in the 1980s. The question in 2023 was: is this another cycle, or is something genuinely different this time?
Something is genuinely different this time.
What Changed
The previous waves of AI were impressive in narrow domains — chess, image recognition, fraud detection — but brittle. They required enormous amounts of labelled training data, specialised expertise to build, and produced systems that could only do one specific thing.
Large Language Models (LLMs) broke that pattern.
GPT-3, and then GPT-4, Claude, Gemini, and Llama demonstrated that a single model, trained on text at massive scale, could write code, summarise documents, answer complex questions, translate languages, generate images (via diffusion models), and hold nuanced conversations — all without being explicitly programmed for any of these tasks.
The shift was from narrow AI to something that felt, for the first time, genuinely general.
What GenAI Actually Changed About Work
The most honest way to describe it: GenAI compressed the distance between idea and output.
A developer who previously spent an afternoon writing boilerplate code now describes what they want and gets a working draft in seconds. A content writer who struggled with a blank page gets an outline, a draft, and five alternatives in minutes. A data analyst who couldn't write Python can now write plain English and get working code back.
This didn't eliminate jobs — it changed what skills were valuable. The ability to prompt effectively, to evaluate AI output critically, to integrate AI tools into workflows became as important as the underlying technical skills.
The New Career Landscape
The pattern from 2009 is repeating — but faster:
- AI/ML Engineers are commanding some of the highest salaries in the industry
- Prompt Engineering emerged as a skill nobody had in 2021 and everyone wants in 2024
- FinOps for AI — managing the runaway costs of inference at scale — became a specialisation
- AI Safety and Ethics roles proliferated at major labs and regulators
The certification economy is booming again. AWS, Google, Microsoft, and Anthropic all have AI-focused credentials. History rhymes.
What Hasn't Changed
Through every wave — recession, certifications, cloud, DevOps, IoT, AI — the engineers who remained valuable shared common traits:
They stayed curious. They didn't wait for their company to train them.
They understood fundamentals. The person who understood why TCP/IP works adapted to cloud networking faster. The person who understood operating systems adapted to containers faster. Fundamentals don't expire.
They built things. Not just read about them. Side projects, experiments, broken deployments at 2am — that's where the real learning happened.
Where We Are Headed
GenAI is not a destination — it's a new baseline. In five years, AI assistance in software development will be as unremarkable as using Stack Overflow is today. The question won't be "do you use AI?" — it'll be "how well do you use it?"
The engineers entering the field now are the luckiest in history. They have tools their predecessors could not have imagined. They also face the most complex, rapidly-shifting landscape the industry has ever produced.
The ones who will thrive are the same ones who always have: curious, adaptable, and unwilling to stop learning.
This post is part of a series tracing the arc of technology careers from the 2009 recession to today. Start from the beginning: 2009: Recession, Resumes, and the Rise of the Training Institute.