Why programming with AI is fun again

GitHub CEO Thomas Dohmke sees a new era of software development with Copilot. A German is the driving force behind the tool that doubled my productivity.

Why programming with AI is fun again
Image: AI-generated illustration

For me, AI is nowhere stronger than in software creation. I am enjoying programming again – and I am more productive than I have ever been. A German is the driving force behind this development: Thomas Dohmke.

What is at stake

Thomas Dohmke has led Microsoft subsidiary GitHub since 2021. Its archive holds a large share of the world’s published software. With the AI tool Copilot, his team brought one of the first mass-market products of the AI revolution to market – and after two years he gave me a stocktake in San Francisco that is increasingly difficult to argue with.

How much AI has already changed coding

When Copilot launched in 2021, 20 per cent of the code created on the platform came from AI. In some programming languages it is now already more than 60 per cent. More than a million developers and 37,000 companies use the tool, including Shopify, Maersk and the consultancy PwC. Dohmke’s thesis: the threshold for programming is dropping so sharply that years of perfecting a programming language may soon become unnecessary. Nvidia CEO Jensen Huang puts it even more bluntly: “Today, everyone is a programmer.”

Why the balance of power could shift

Until now, elite software work was the domain of tech giants that could afford armies of highly paid developers. AI assistants are shifting that advantage. A mid-sized company can now take on projects that were out of reach yesterday. Daniel Szabo, CEO of Körber Digital, told me: “The next Google could be a German mid-sized company.” In an internal analysis, 82 per cent of Körber’s programmers reported that AI assistants made them faster – while accepting only 28 per cent of AI suggestions. The technology is powerful, but not infallible.

Why jobs will remain – and matter more

The fear that AI will make developers redundant misses the opposite dynamic. Economists call it the Jevons paradox: when a resource becomes more efficient to use, demand often rises rather than falls. The easier it is to build software, the more companies will want to build it – and the more important the people who steer, review and translate the results into viable products will become. Dohmke’s warning nonetheless: “The technology offers enormous opportunities, but Germany is simply too slow to move.” Anyone who wants to compete at world-class level cannot afford to hesitate.

I wrote the full piece together with Thomas Jahn for Handelsblatt.

More on Artificial Intelligence

All stories