You’re reading AI at the Top - Our new series on Sundays, where we share how leading companies and their executives use AI in business.
Not long ago, Vibe Coding was all over the internet.
We have seen AI-curious people use tools like GitHub Copilot, Lovable, Bolt, Cursor, Claude Code, etc., to grow from Idea → App → Revenue in a flash.
With “Idea to app in seconds” being the new normal, we bring GitHub CEO Thomas Dohmke’s views on how you can build effectively with AI.
If you know Thomas, but not his journey, here is a quick timeline of how he ended up as GitHub CEO:
In 2010, he co-founded Codenauts, an agency that built mobile apps for 30+ German enterprises.
When management became a roadblock for handling multiple clients, he co-founded The HockeyApp in 2011 to solve beta app distribution, crash reporting, analytics, etc., for mobile app developers.
By 2014, HockeyApp was a global DevOps solution for mobile devs. In the same year, Microsoft acquired HockeyApp.
At Microsoft, he led the integration of HockeyApp into Visual Studio App and managed various dev tools.
After Microsoft acquired GitHub in 2018, he transitioned to GitHub as VP of strategic programs, then CPO, and as CEO since 2021, helping the company build Copilot, Copilot X, and GitHub models.
In today’s AI at the top, we will share Thomas’s views on:
Why should you vibe code?
How to vibe code effectively?
Thomas’ favourite AI tools
Why should you vibe code? Thomas thinks…
The gap between idea and app has criminally reduced.
You can come up with an idea in the morning shower and demo the solution to your friends at dinner.
Instead of telling what’s on your mind, you can vibe code a prototype and show what you mean.
What does this mean?
It demonstrates your commitment, clear thinking, and the ability to shape ideas into reality.
In fact, large companies like Google changed their ways to measure clear thinking:
Also, the roles are overlapping. Workscopes are expanding.
“I like the story of this product manager who vibe coded a feature he felt was important. Later, the feature made sense business-wise, so the dev team shipped it.” Thomas shared.
That’s the power of building with AI. You never have to wait to act.
In the AI at the Top series, we interview executives and share how leaders use AI to improve business efficiency. If you are one such leader, reach out to us.
Now that we know vibe coding is cool, how do we get good at it?
Thomas thinks there are two layers to vibe coding.
Machine language layer is deterministic. Any person who wants to build a feature should write the code in the same way for it to function. More or less.
Human language layer is non-deterministic. Three people building the same feature prompt AI differently from each other, leading to outputs that are poles apart.
So of course, it’s basic to get super good at prompting.
But Thomas encourages you to learn Systems Thinking - the ability to break down a large solution into small features and know how they’re connected.
AI-driven companies like Netflix and BMW use systems thinking to map out how every new feature tweak will ripple across their products, so nothing breaks with new changes.
“Can’t I just ask AI to do it for me?”
Yes you can. But then you will have to judge if it takes your idea in the right direction or not.
Think of systems thinking as a taste to engineer apps with AI.
GitHub CEO’s favourite AI tools
GitHub Copilot: Code completion, debugging, learning new programming frameworks
GitHub Copilot X: Natural language documentation search, code, pull request reviews
Claude: Code analysis for workflows
Gemini: Multi-language or cross-platform coding tasks
Open Source LLMS (StartCoder, Code LLama): Custom development needs
OpenAI GPT-4: Regular questions that make him curious; replacement of traditional search
Granola AI: To transcribe meetings
Image Generation tools like Midjourney: “I am not creative, so I keep using image generation models for my presentations.”
Takeaways and TL;DR:
If you have read this far, you know what to do next:
Not ideas, but create prototypes. This is your chance to break the cliché quote, “Ideas are worth nothing without execution.”
Develop systems thinking. Learn to break down complex solutions into small features, and how you connect them for the best output.
We always say: Models evolve fast. Don’t stick to one tool; play around with them and use what works best at the time.
…and that’s our time.
Reply with your vibe-coded projects; we’d love to know what our r leaders are cooking.
To shaping ideas with AI! 🤖
Sources:
Stay curious, {{first_name | leaders}}
PS. If you missed yesterday’s issue, you can find it here.