GitHub COO Kyle Daigle on 'vibe coding' and why one should still learn to code in the AI era

GitHub COO Kyle Daigle

Learning to code remains an essential skill for the future, GitHub COO Kyle Daigle told Moneycontrol in an interview, even as artificial intelligence (AI) tools are becoming increasingly adept at writing code themselves.

"Coding is problem solving. Learning how to solve problems is a skill that has been necessary and true since the dawn of time and will continue to be true in the age of AI" Daigle said.

His remarks come amid an ongoing debate over whether one should learn to code or not. Replit CEO Amjad Masad recently said that it would be a "waste of time," while Anthropic CEO Dario Amodei stated that AI could be 'writing essentially all of the code' in 12 months, indicating that traditional programming skills may no longer be necessary to build software.

AI pioneer Andrew Ng, however, disagrees, arguing that as coding becomes easier, more people should code. He believes that people should learn enough coding to effectively use AI-assisted coding tools.

Daigle said in the interview that while AI will increasingly handle the repetitive and routine parts of coding, having coding skills help people understand how to solve the problem.

"I believe learning to code is an essential skill...just like we still teach everyone how to do mathematics, even though we all just use calculators...You can then utilise tools to help you solve the problem faster" he said.

Daigle was also enthusiastic about the ‘vibe coding’ phenomenon, a term coined by OpenAI co-founder Andrej Karpathy to describe a new form of programming that heavily relies on AI coding tools to generate code from natural language prompts describing an idea, rather than writing code line by line.

"I want coding to be more accessible for everyone...I think vibe coding is just watching the magic that every professional developer has experienced, just without all the friction" Daigle said "Letting people be excited by software has been the core of GitHub's mission for 17 years"

He said that GitHub also helps vibe coders secure their apps through its code security tools, which detect vulnerabilities that can then be resolved using GitHub Copilot, the company's flagship code completion tool.

GitHub, which was acquired by Microsoft for $7.5 billion in 2018, has been infusing AI capabilities with a slew of new products and feature enhancements across every step of the developer lifecycle in the past few years.

Last week, GitHub rolled out Agent Mode to all users of VS Code(Visual Studio Code), an integrated development environment (IDE) developed by Microsoft. The feature, first previewed in February, enables developers to turn their ideas into code through simple prompts.

In addition, GitHub also announced the public availability of Copilot code review agent that helps speed up code reviews, and the next edit suggestions feature that auto-recommends the next code change based on previous edits.

In February, GitHub had also unveiled its plans for an autonomous agent called "Project Padawan", designed to independently handle entire tasks at the developer’s direction. The feature will enable developers to assign issues to Copilot, let the AI complete the task autonomously, and return later to review its work

Daigle said such features also allow professional developers to join in on the vibe coding phenomenon.

"I think sometimes people talk about vibe coders as only those who don't know how to code. But I think it's just about 'I have an idea, and I want it to exist'. I don't necessarily want to type out every word, every syntax, or every method" he said.

As AI models get increasingly powerful, Daigle believes that more innovation needs to happen on how people interact with them, moving beyond a chat interface.

"We're in the adolescence of AI UX (User experience), and the future will be agentic, and that means far more autonomous for doing the tasks before I need to ask for them...We need startups and enterprises to investigate new user experiences" he said.

Daigle believes context-aware AI coding will be the next big area, helping transform the platform from a one-size-fits-most tool into a 'built-for-you' AI tool tailored to each person’s project, team, and needs.

"When we look across all the information we have around the code - issues, pull requests, teams, and how they engage together - we're still talking about code within an IDE...As we bring in Agent Mode and Project Padawan, if we can expose things like the code, issues, rejected suggestions, pull requests that didn't go through, team structure, seniority and expertise of the teams, we can truly build AI, particularly AI agents, that match our needs without us having to ask them to do work" he said.