AI-Powered Coding: GitHub CEO Thomas Dohmke on the Future of Software Development
Thomas Dohmke, CEO of GitHub, joins Dr. Jeanne Fredericks at Africa Tech Festival 2024 to discuss AI-powered coding, the evolution of Copilot, and how AI is democratising software development across the world.
The AI Revolution in Software Development
Note: The below article has been created using a transcript of the session video.
AI is transforming software development at an unprecedented pace. At Africa Tech Festival 2024, Thomas Dohmke, CEO of GitHub, shared how AI-powered networks, Copilot, and multi-modal AI models are changing the way developers write, test, and deploy code.
"Four years ago, Copilot was just an idea," Dohmke recalled. "Now, it's writing 25% of code in projects, and AI is changing the world."
The conversation highlighted three key themes shaping the future of software engineering:
- The Origin and Impact of AI-Powered Coding
- Multi-Modal AI & AI-Native Workflows
- How AI is Democratising Software Development
How AI-Powered Coding is Changing Development
GitHub’s Copilot was born during the pandemic lockdowns of 2020, when Dohmke’s team tested OpenAI’s GPT-3 model for coding tasks.
"We asked it to write a method in JavaScript to calculate a prime number, and it did it—perfect syntax, logic, everything. That moment was mind-blowing," Dohmke shared.
From there, GitHub embedded AI-powered code suggestions directly into developers’ editors, helping them write code faster, reduce errors, and boost productivity.
"AI doesn’t replace developers—it supercharges them. Even a 10% efficiency gain means four extra hours of productivity per week. That’s game-changing."
Today, Copilot is writing one-quarter of all code in GitHub projects, with AI playing a larger role in debugging, testing, and deployment.
Multi-Modal AI & The Future of AI-Native Workflows
As AI models evolve, developers no longer rely on just one AI system.
"Developers need flexibility—so we introduced multi-modal AI choice," said Dohmke.
This means that GitHub Copilot now integrates multiple AI models, including:
- OpenAI's GPT models
- Google’s Gemini
- Anthropic’s Claude
- Open-source AI models
"No single AI model is best for everything," Dohmke explained. "Developers should be able to choose the right AI for the right task—whether it's writing, testing, or debugging."
Even more revolutionary? The rise of AI-native development.
"Soon, you won’t even need a laptop to code. AI-native tools will let you build apps directly on your phone, just by describing what you want in natural language."
With Copilot Workspace, developers can:
- Turn ideas into working code without manually checking out repositories
- Use AI agents to refine, debug, and improve software
- Code in any human language, including Afrikaans, Spanish, and Hindi
"The future of coding is about working alongside AI—not just using it as a tool, but as a true development partner," Dohmke emphasised.
AI is Democratising Software Development
GitHub has an ambitious goal: 1 billion developers worldwide.
"Software is just digital literacy—every child should learn it, just like math or reading," Dohmke said.
Thanks to AI-powered coding, learning to program is easier than ever:
- Copilot works in multiple languages, so kids can learn to code in their native tongue.
- AI has infinite patience—unlike teachers or coworkers, it never criticises, just keeps improving answers.
- Anyone with a smartphone can start coding—no expensive tools or degrees required.
"AI doesn’t just make existing developers better—it opens the door for millions of new developers who never thought coding was for them," Dohmke concluded.