Three Cambrian Moments
Software Engineering careers in the age of AI
It’s the third time in my career I’m witnessing a Cambrian moment in the software industry: first the birth of the open Web (thank you TimBL), then the open availability of large-scale data processing systems (thank you DougC), now the broad availability of large-language models used for GenAI (thank you Google for the T in GPT).
I’ve been lucky enough to have access to front row seats every time, my modest self attributing this to sheer contemporary luck. More honestly, each time I also felt a bit like Statler or Waldorf sitting in the balcony, intently watching the theatre play unfold below, the audience booing and then cheering The Greatest Show On Earth (1952), or more recently The Greatest Showman (2017).
What is truly amazing is the compounding velocity of these three moments: each revolution builds on the other previous one, happens faster and with more surgical precision.
Inventions to improve scientific knowledge sharing (the Web), to number-crunch human behaviour into prediction models (Hadoop), have now transformed into general-purpose and multi-modal AI models to mimic, even replace our human neural networks with broad applicability: recognition, summarisation, classification and labelling, translation, generation of near-human-crafted prose and visuals, … now firmly entering the era of specialised software coding models.
The GenAI inmates are running the asylum
The importance of the advent of serious software coding models such as Anthropic’s Opus 4 packaged in Claude Code cannot be stressed enough. It’s a category-defining moment in time, equipping serious software engineers with superpowers to 10x productivity: I pity the fresh-graduating junior software engineer, or worse the self-taught “a job in IT because it’s cool + money” opportunists.
Carbon-based (human authored) code is being replaced with electron-generated. The next generation of tools, frameworks and models will be 10% human inspiration, 90% electron transpiration. And with agentic systems popping up everywhere, the future of software is going to be very different than the past 50 years.
There are risks to be controlled and guardrails to be set, and looking at humanity’s unruly history it’s unfortunate but likely that the next generation of Zippe-type centrifuges will instead be rogue-state-operated unbounded AI. We will need an IAEA for AI eventually, and it will have to employ AI to combat rogue AI.
But I’m getting carried away here.
Impact on Software Engineering careers
The impact of AI on Software Engineering careers can also not be overstated. The educational system has barely adjusted to accommodate some modest inclusion of AI in its curricula, after a period of pushback due to plagiarism obsession. True, creative thinking cannot be born out of next-best-word-prediction models, and our carbon-based neural networks aka our brain is still uncannily superior (most of the time) in relevancy in precision and recall, but we will have a hard time to beat the computers in the efficiency race.
It’s too bad AI models require nuclear plants to run, but there’s an equally serious shortage of Einsteins throughout the human population. I’m pretty sure humans working with AI will eventually solve the nuclear fission clean energy problem faster than what we can (want to?) do without.
All this just to say that Houston, we have a problem. Letting graduate junior software engineers that are not fully AI-equipped makes limited sense anymore, as their usual avenue of employment will be replaced by a 200$ subscription to a software coding model.
The Cambrian moment of AI will have a serious impact on the careers in the software industry itself. Craftsmanship is slowly but surely getting replaced by good-enough coding assistance. Products are turning into services, with agents stripping away the need for UX design. It’s quite unclear but certainly worrying that the window of predictability of a career in software engineering is getting smaller than ever.
For years, software engineering schools have tried to step away from a heavy focus on CS fundamentals (data structures and algorithms anyone?) in favour of more practical, applied learning: applied software engineering, processes, frameworks, tooling… the latter exactly where AI shines. We will need to reverse this trend and double-down on smart math skills, architecture, data… with coding to be handled by the computer.
It’s not broken until we fix it
Don’t you worry: I strictly relied on my own carbon NNs to write this post:
Where we humans still excel at are value- and belief-based systems,
optimism, true creativity, patience-to-nurture-the-problem, connection.
We will use AI for its benefits and to our benefit, if we embrace it at the right time. That time is now, and it’s going to be The Greatest Show On Earth!
With love,
Steven.