Thomas Dohmke, CEO of Github It says “the key to winning” AI coding tools Rather than relying entirely on automated agents, it maintains the ability to manually quickly modify AI-generated code. Speaking about “Mad Podcast with Matt Turku,” Domeke emphasized that developers need the flexibility to move seamlessly between AI support and practical coding.The CEO outlined an ideal workflow where AI agents can write code and send pull requests, but developers retain the ability to use existing ones to make rapid changes Programming skills. This approach prevents minutes of productivity traps explaining simple changes in natural language when direct coding can complete the same task in seconds.“The worst option is to try and find a way to explain feedback or prompts in natural language to explain what you already need to do in programming language,” Dohmke explained. Such a scenario could turn a 3-second task into a 3-minute trial, and ultimately reduce it rather than increasing productivity.
Startups cannot rely solely on “vibe coding.”
Dohmke also addressed trends.”Vibe coding“The phenomenon, a term that explains Openai's co-founder Andrej Karpathy relies heavily on AI for code generation. In a recent Q&A at Station F in Paris, Github CEO warned that startups cannot scale with atmospheric coding alone.He argued that non-technical founders struggle to build sustainable companies without skilled developers, saying “we can't build complex systems to justify the next round.” The key is to enable developers to choose the best approach for each situation. This is whether to leverage AI agents for maximum ROI or handle tasks directly.
