2025 was a strange year for software engineers. Equal parts exhilarating and terrifying.
We can now build things four or five times faster than before. Features that took weeks or months now take days. Prototypes that took days now take hours. The bottleneck has shifted from “can we build this?” to “should we build this?”
This is not a productivity boost. This is industrialization.
The Factory Floor
Every industry goes through industrialization eventually. What was once craft becomes production. What was once rare becomes abundant. And when that happens, the rules change.
Consider this: a quarter of startups in Y Combinator’s Winter 2025 batch have codebases that are 95% or more AI-generated. At Google, AI already writes more than a quarter of all new code.
The transformation is global but uneven. In Africa, McKinsey found that more than 40% of institutions have either experimented with or implemented generative AI. The continent is home to roughly 2,400 companies specialising in AI - about 40% of which are startups. Africa’s AI market is projected to grow from $4.5 billion to $16.5 billion by 2030, and the IFC estimates that 230 million jobs in Sub-Saharan Africa will require digital skills by 2030.
The textile industry went through this. Publishing went through this. Now software is going through it - everywhere, at different speeds.
When textiles industrialized, more people could afford clothes. When publishing industrialized, more people could read books. Maybe when software industrializes, more ideas can become real.
But here’s the uncomfortable part: quality used to be the differentiator. Now quantity is.
When Veracode tested over 100 AI models on real coding tasks, 45% of the code they produced contained known security vulnerabilities. GitClear’s analysis of 211 million changed lines of code shows copy-pasted code rising sharply while refactoring collapses. The craft is becoming a factory - with all the trade-offs that implies.
What Happens to Jobs?
For years, the advice was clear: specialize. Pick a lane. Become the React expert. The Kubernetes wizard. The machine learning guru. Depth was the moat.
But here’s the thing about AI - it’s already deeper than most of us in most things. It has read every Stack Overflow answer, every GitHub repo, every tutorial. It doesn’t forget. It doesn’t get rusty.
The job market tells a contradictory story. The World Economic Forum projects AI and machine learning specialists among the fastest-growing roles through 2030, with demand growing around 40%. Hiring is hot for engineers in AI/ML, MLOps, and cloud infrastructure. But generic full-stack roles are cooling.
The tension is visible in Africa’s growing tech ecosystem. Google and Accenture counted over 716,000 professional developers across the continent in 2021 - a pool growing every year, yet still far short of what the continent’s own digital transformation demands. Global recruiters say niche roles - site reliability engineering, cloud and AI systems - offer the best opportunities.
But specialization alone isn’t enough. One recruiter described senior African developers taking mid-level roles abroad because their experience doesn’t include large-scale AI infrastructure. Those who break through do so by “stacking scarce skills” - combining specializations in ways that create unique value.
Meanwhile, two-thirds of African companies are reskilling staff with AI tracks while using automation to fill gaps where human expertise is lacking. The pattern seems to be: depth on demand, but breadth as the foundation.
Here’s what concerns me: 52% of developers say AI has positively affected their productivity. But when actually tested in a controlled study, experienced developers took 19% longer on tasks with AI than without. The extra time came from checking, debugging, and fixing AI-generated code.
And yet - they believed they worked 20% faster.
The perception and the reality don’t match. We’re building careers on assumptions we haven’t verified.
What Happens to Teaching?
If the skills that matter are shifting, should we teach software engineers differently?
We’ve built entire curricula around specialization. Data structures and algorithms, yes, but then branch early - mobile or web, frontend or backend, AI or infrastructure. The assumption was that depth creates job security.
Something is already shifting.
Kenya’s National AI Strategy (2025-2030) makes education a priority by embedding AI literacy across primary, secondary, and higher education. Tunisia’s AI Roadmap focuses on awareness, skills, and infrastructure. Morocco allocated 50 million dirhams to the Al-Khawarizmi programme to fund 45 AI research projects, and has since created its first national research institute dedicated to artificial intelligence.
Rwanda’s New Generation Academy - the country’s first accredited institution for software programming and embedded systems - opened its doors in late 2025. In Nigeria, Data Science Nigeria provides intensive training across both Anglophone and Francophone regions. And according to one analysis, South Africa, Kenya, and Nigeria comprise 62% of recognized AI degree programs in Africa.
But challenges remain stark. Kenya’s own AI strategy lists the baseline plainly: 65% of schools lack internet connectivity. In other parts of Africa, researchers note that universities “are not really ready to adopt AI policies due to lack of material, curriculum updating, and teachers’ training.”
Meanwhile, Columbia Engineering in the US is reworking its programming curriculum. Now students spend time assessing the quality of AI-generated code, learning to evaluate correctness, efficiency, and design. The focus is shifting from using AI to write code toward using AI to understand and improve it.
Notice what’s happening across all these contexts? The emphasis is moving from “learn this specific technology” to “learn how to learn, evaluate, and adapt.”
Here’s the thing: this was always what education should have been about.
For decades, “critical thinking” and “problem-solving” were treated as niche concerns - soft skills for progressive educators to champion while the real curriculum focused on pouring information into students’ heads. Across Africa, critics have long pointed out that education systems inherited from colonial times were designed to train office workers, not thinkers. Students memorize vast amounts of information, regurgitate it for exams, and forget most of it afterward. The curriculum produces graduates who hold little real-world value in solving the continent’s pressing challenges.
This wasn’t a secret. The African Union made 2024 the Year of Education, pushing for curriculum reforms that foster critical thinking and creativity. Kenya’s new Competency-Based Curriculum explicitly prioritizes problem-solving, imagination, and critical thinking over exam success. Rwanda’s African Leadership University offers a “Global Challenges” degree built entirely around solving real-world problems.
The shift was already happening. But AI has put a spotlight on how urgent it really is.
When machines can memorize everything and retrieve it instantly, what’s the point of training humans to do the same? The skills that were always essential - thinking critically, solving novel problems, connecting ideas across domains - are now the only skills that matter.
The World Economic Forum identifies creative thinking, flexibility, resilience, and lifelong learning as skills rising sharply through 2030. Not framework expertise. Not language mastery. Adaptability.
The Next Generation
What should we tell students starting their software engineering journey today?
I don’t have a definitive answer. But the old playbook (that already needed revision ages ago) is now completely obsolete.
The students entering university now will graduate into a world where AI writes most of the boilerplate, handles most of the syntax, and knows most of the APIs. What’s left for humans?
Maybe it’s the parts that don’t fit in a prompt window. Understanding context. Asking the right questions. Knowing what to build, not just how to build it. Connecting technical decisions to business outcomes. Navigating ambiguity.
Maybe the best preparation isn’t learning to write code faster - it’s learning to think about systems, products, and people. The kind of breadth that lets you evaluate what the machine produces, not just consume it.
From Lagos to San Francisco, from Nairobi to Berlin, from Casablanca to Singapore - engineers are wrestling with the same questions. What skills matter now? How do we prepare for jobs that don’t exist yet? What does it mean to build software when the building is increasingly automated?
The industrialization of software is not coming. It’s here.
The question is what we do with it.