GitHub CEO predicts the future of programming (AI)

I watched it this weekend. interview Thomas Dohmk, CEO of GitHub, was taken from him by Matthew Berman, a prominent AI blogger. The conversation took place after MS Build, where GitHub announced open source for its chat inside VSCode.GitHub Copilot Chat" The main thoughts of the interview below

1. The Influence of GPT and the Beginning of Copilot GitHub’s CEO admits he initially doubted GPT’s performance, but was impressed by the results: GPT changed the way we develop software forever. Copilot became the first mass-produced tool that completes code per user, and it turned out to be extremely popular. 2. Statistics and user experience Copilot writes 25% of code in files where it is enabled and user satisfaction is very high (NPS — 72). Integration with editors (VS Code) Code completion has become the entry point of AI into the daily work of the developer. 3. Education and importance of programming Programming remains an important skill for understanding the modern world and must be taught to children and adults. The basics of computer science and the ability to read code are the universal literacy of the future. 4. The role and responsibility of the developer Even with the advent of copilot agents, an engineer must verify and confirm changes to avoid errors and safety issues. It’s important to understand what an AI agent does and not lose control of the code. 5. Industry Evolution and Open Source Open source has become the industry standard, and Copilot is now available as open source for community integration and refinement. (This is an extension for chat.). GitHub and Microsoft are betting on using multiple AI models for different tasks, rather than one universal one. (One for code completion due to latency requirements, another for agency workflow, where latency is not as important and you can use a smarter and slower model.) 6. The Future of Development: Agents and Micro Applications An interesting idea is that in the future, personal agents can create mini-applications “on the fly” for specific tasks, and the operating system will be hidden behind the AI interface. Example: Automation of family tasks (managing children's pocket money) with Copilot and personalized applications 7. Vibe Coding and the Limitations of AI Vibe Coding is an approach in which AI helps to quickly turn ideas into prototypes, but for serious tasks you need a review, security and standards. AI-agents are limited in the amount of code processed, but in the future will be able to work with large projects, and then vibcoding will become even easier. 8. Integration and ecosystem of AI agents An ecosystem of personal AI agents integrated with the work and personal life of the user is expected. Work-related knowledge will belong to the company, and personal knowledge will remain with the user. In some ways, this concept resembles a series.Separation" which I have not seen, but which I have heard of." 9. Impact on the labour market and new professions AI automates routine tasks, but opens up new professions and opportunities, even for those who do not speak English. As in past technological revolutions, some professions will disappear, but new roles and specializations will emerge. 10. Optimism for AI Thomas is optimistic that AI will not replace people, but will help solve more tasks, increase productivity and create new opportunities for developers.

Overall, the interview shows how fast AI-assisted programming is evolving and what role GitHub plays as an industry leader. AI does not replace programmers, but makes them more productive, allowing them to focus on the creative and strategic aspects of development.

#AI #Software #Engineering #Architecture #Agents #Math #Software #ML #Leadership