The hottest new programming language is English. Andrej Karpathy, OpenAI co-founder and former Tesla AI director, said this in 2023. Two years later, 41% of all global code is AI-generated and “vibe coding” is Collins Dictionary’s Word of the Year. The implications for South African business leaders are not theoretical. They are immediate.
The most powerful programming language in the world is not Python, not Java, and certainly not the binary sequences that preoccupied the first generation of computer scientists. It is the language you are reading this article in.
That shift − from machine-speak to human-speak − took 70 years to complete. And in 2025 and 2026, it stopped being a theoretical argument and became a measurable, data-verified reality with names, tools and market valuations attached to it.
In my years advising enterprise leaders across telecommunications, financial services and emerging technology in Africa and beyond, the most consistent gap I encounter is not a shortage of enthusiasm for AI. It is a fundamental misunderstanding of what this shift actually makes possible − and what it demands of leaders whose organisations are not yet inside it.
Seventy years in eight sentences
Programming languages began as machine code, raw binary that processors executed but humans could barely read.
Assembly added mnemonics. FORTRAN and COBOL introduced English-like syntax. Pascal and C brought structured discipline. Object-Oriented Programming modelled the real world as objects and actions. Java and Python made software internet-native. Go and Rust made it cloud-native.
The 45% security failure rate in AI-generated code is not a theoretical risk; it is a board-level liability in a jurisdiction with real enforcement teeth.
Each generation moved closer to human expression and further from machine formality. The arrival of large language models (LLMs) was not a disruption of this trajectory. It was its logical conclusion.
Vibe coding: The consumer face of the revolution
On 2 February 2025, Andrej Karpathy posted six words that redefined how the technology industry thinks about building software: “Fully give in to the vibes.”
He was describing vibe coding, a development approach where you express what you want in natural language and let AI generate the implementation. Within months, the practice moved from viral tweet to mainstream workflow. Collins Dictionary named it Word of the Year for 2025. Searches for the term reportedly jumped 6 700% in a single quarter (Exploding Topics, 2025).
The numbers are arresting. By 2025, 84% of developers globally were using or planning to use AI coding tools (Stack Overflow Developer Survey 2025). AI-generated code now accounts for 41% of all code written worldwide. Among Y Combinator’s Winter 2025 cohort, 21% of companies had codebases that were over 90% AI-generated. The vibe coding platform market was valued at $4.7 billion in 2025, projected to reach $12.3 billion by 2027. This is not an experimental workflow. It is mainstream engineering practice.
Vibe coding shifts your role from writing code to directing an AI. You focus on the product. The AI focuses on the syntax. The barrier between intent and implementation has, for the first time, dissolved.
AI coding assistants: The enterprise-grade layer
Below the consumer energy of vibe coding sits the more governed, enterprise-calibrated layer: AI coding assistants.
GitHub Copilot leads the market with over 20 million users as of mid-2025, a 400% year-on-year increase − and 90% adoption across Fortune 100 companies.
A controlled study across 4 800 developers found tasks completed 55.8% faster using Copilot. The AI coding assistant market reached $7.37 billion in 2025, projected to grow at 27.1% CAGR through 2032. Gartner’s 2025 Magic Quadrant confirms GitHub as category leader, alongside Amazon, Google Cloud and Cognition.
But the governance tension is real and must be named. CodeRabbit’s analysis found AI-coauthored pull requests show 1.7 times more issues than human-written code. Veracode’s 2025 GenAI Code Security Report found 45% of AI-generated code fails security tests.
Gartner’s own survey found 42% of engineering staff report productivity gains of only 1% to 10%, a reminder that tooling alone does not deliver outcomes. Governance, review and architectural judgement remain irreducibly human.
What this convergence means for South African business
The convergence of vibe coding and AI coding assistants has specific implications for South African organisations. The technical barrier between business intent and working software has collapsed, but only for organisations that know how to direct AI purposefully.
A business analyst can now prototype a data dashboard. A compliance officer can draft a workflow automation. A product owner can build a functional MVP without writing a single line of code. These are present realities and South African enterprises that have not updated their operating model to reflect them are already behind.
POPIA adds a dimension that is uniquely critical in our market. Every AI-assisted workflow that generates, processes, or stores South African customer data carries a data governance obligation.
Vibe-coded and Copilot-assisted software that ships without rigorous security review is a POPIA exposure waiting to materialise. The 45% security failure rate in AI-generated code is not a theoretical risk; it is a board-level liability in a jurisdiction with real enforcement teeth.
The horizon: From assistants to agents
Prompt engineering, the craft of constructing AI inputs that reliably produce high-quality, auditable outputs, is already a formal discipline. Agentic AI, where systems act autonomously across multi-step workflows, is the next layer: McKinsey’s State of AI 2025 survey found 62% of organisations are already experimenting with autonomous AI coding agents.
South African organisations that build prompt literacy and governance infrastructure now will be positioned to deploy agentic capability responsibly as it matures, rather than scrambling to govern it retrospectively.
The language dimension remains Africa’s most underappreciated opportunity. The LLMs powering vibe coding and AI coding assistants were trained overwhelmingly on English and European-language data.
Building AI systems that reason fluently in isiZulu, Sesotho, Xhosa and Afrikaans is not a niche concern. It is the AI capability gap that will determine which organisations can genuinely serve South African markets, not just serve the English-speaking segment of them.
The skill that now defines technical leadership
The 70-year journey from binary code to natural language is complete. But the lesson is more nuanced than “anyone can code now”.
The data shows the organisations extracting real value from vibe coding and AI assistants are those that combine AI speed with human judgement, clear problem definition, disciplined governance and critical evaluation of AI outputs. These are not soft skills. They are engineering requirements, and they live in the business function, not the IT department.
My question to every South African leader reading this: your organisation almost certainly has people with deep domain knowledge, strategic clarity and the ability to communicate with precision. Are you treating those people as your most valuable AI-era technical asset?
Because the research is unambiguous: the productivity gains go to the teams that direct AI with expertise, not simply to the teams that use it.
* Eugene Perumal is a strategy and architecture principal with over 20 years' experience in enterprise technology across telecoms and financial services, including senior roles at Vodacom Group and Absa Group. He holds Master’s degrees and certifications in enterprise architecture, AI governance, cloud and analytics. He writes on enterprise AI strategy, ROI measurement and the shift to agentic AI deployment.
Share
