Podcast Summary: The MAD Podcast with Matt Turck
Episode: GitHub CEO: The AI Coding Gold Rush, Vibe Coding & Cursor
Date: June 12, 2025
Guest: Thomas Domke, CEO of GitHub
Host: Matt Turck
Episode Overview
This episode features an in-depth conversation with Thomas Domke, CEO of GitHub, about the transformative impact of AI on the world of coding and development. The discussion spans the strategic acquisition of GitHub by Microsoft, the pioneering development of GitHub Copilot, the frenetic landscape of generative AI for coding, the concept and future of agentic coding, and the evolving balance between competition and collaboration among major AI and developer tools companies. Both coders and non-coders will gain valuable insights about where software development is heading and what it means for the future of work, SaaS, and the developer ecosystem.
Key Discussion Points & Insights
1. The Vision and Strategy Behind Microsoft’s Acquisition of GitHub
- Historical Context
- Microsoft acquired GitHub in 2018 for $7.5B, a price Domke now calls “cheap” by 2025 standards. The primary motivation: to solidify Microsoft’s identity as a developer-focused, platform-centric company and fully embrace open source.
“I think Microsoft has always seen itself as a developer tools company... to be successful with a platform, you need to have developers on your platform that build apps, services, products on top of it, but also build an ecosystem.” (03:12) – Thomas Domke
- Microsoft acquired GitHub in 2018 for $7.5B, a price Domke now calls “cheap” by 2025 standards. The primary motivation: to solidify Microsoft’s identity as a developer-focused, platform-centric company and fully embrace open source.
- Acquisition Best Practices
- Domke highlights the principle of acceleration: acquirers should uplift target companies as a VC would an investment.
“The buyer should accelerate the target company... when you’re acquiring companies, I think you should have that same mindset, buying a company to accelerate them so they increase their value.” (04:20) – Thomas Domke
- Domke highlights the principle of acceleration: acquirers should uplift target companies as a VC would an investment.
- Synergy with Azure
- GitHub significantly drives Azure’s developer pipeline and revenue by onboarding massive numbers of developers who might become Azure users.
“GitHub revenue became part of the Azure KPI... between 2017 and 2024 revenue went 10x right? And... Copilot is a big part of that cloud...” (07:54) – Thomas Domke
- GitHub significantly drives Azure’s developer pipeline and revenue by onboarding massive numbers of developers who might become Azure users.
2. How GitHub Became a Generative AI Pioneer
- Early AI Bets & Copilot’s Origin
- GitHub Copilot launched in 2021, well before the GenAI craze triggered by ChatGPT. Microsoft’s early OpenAI investment gave GitHub advance access to GPT-3 and informed the product vision.
“We ask it to write methods like prime number detection... and that ultimately gave us the confidence we can build a product here. So this is 2020, two years and a bit before ChatGPT.” (12:26) – Thomas Domke
- GitHub Copilot launched in 2021, well before the GenAI craze triggered by ChatGPT. Microsoft’s early OpenAI investment gave GitHub advance access to GPT-3 and informed the product vision.
- The Real Value: Workflow, Not Just AI
- The key insight was maintaining developer “flow state”—removing the need for context-switching by putting multi-line code suggestion directly into the IDE.
“The magic of Copilot... wasn’t really about AI... it was about giving you a feature in your editor that keeps you in the flow state and makes you more productive and ultimately more happy.” (18:00) – Thomas Domke
- The key insight was maintaining developer “flow state”—removing the need for context-switching by putting multi-line code suggestion directly into the IDE.
3. Demystifying GitHub Copilot and Modern AI Coding Tools
- For Non-Developers: What is Copilot?
- Copilot turns natural language task descriptions (from Jira, a manager, etc.) into code directly in your coding editor—cutting out time spent searching Stack Overflow, etc.
“You want to stay in that [flow] state as long as possible... Copilot does all of that without that context switch.” (17:28) – Thomas Domke
- Copilot turns natural language task descriptions (from Jira, a manager, etc.) into code directly in your coding editor—cutting out time spent searching Stack Overflow, etc.
- For Developers: How It Works Day-to-Day
- Copilot offers multi-line autocompletion, bug fixes, and more—from simple suggestions to agentic workflows that can automate full tasks or refactor codebases.
4. Rapid Evolution: GitHub Models and Model Choice
- Model Flexibility and Developer Choice
- GitHub now lets developers choose their underlying AI models for Copilot, including OpenAI, Anthropic, Google, and even bring-your-own (BYO) models—because there will never be “one model that rules them all.”
“We believe the same is true for models. There is never going to be one single model that rules them all. But the world of software development is just way too broad...” (23:10) – Thomas Domke
- GitHub now lets developers choose their underlying AI models for Copilot, including OpenAI, Anthropic, Google, and even bring-your-own (BYO) models—because there will never be “one model that rules them all.”
- Model Innovation Outpacing Fine-Tuning
- Fine-tuning on enterprise codebases is less practical, says Domke, when models improve so rapidly and code bases are heterogeneous. Modern agents instead retrieve and use contextual information dynamically.
“What customers really mean [by fine-tuning] is that the agent understands the code base... But that’s not only the code that was relevant during the fine tuning process...” (26:11) – Thomas Domke
- Fine-tuning on enterprise codebases is less practical, says Domke, when models improve so rapidly and code bases are heterogeneous. Modern agents instead retrieve and use contextual information dynamically.
5. The “Political Economy” of AI Coding: Competitors and the Ecosystem
- Explosion of the Coding AI Market
- The space is crowded and hyperactive, from hyperscalers (Microsoft, Google) to AI labs (OpenAI, Anthropic, Mistral) and startups like Cursor, which recently reached $500M ARR and a $9-10B valuation.
“I think we should recognize how amazing it is that a developer tools company is the fastest growing startup of all time.” (30:15) – Thomas Domke
- The space is crowded and hyperactive, from hyperscalers (Microsoft, Google) to AI labs (OpenAI, Anthropic, Mistral) and startups like Cursor, which recently reached $500M ARR and a $9-10B valuation.
- Market Segments Domke Sees:
- Bootstrapping/Greenfield Builders (AI generates from scratch)
- IDEs/Mainstream Dev Tools
- AI Model Companies
- Agentic Tools for Existing Code (Brownfield)
- Universal Need: Meet Developers Where They Are
- Success = integrating into existing workflows and providing flexibility—never forcing radical, abrupt changes to developer behavior.
6. Competition Meets Collaboration
- Microsoft's "Compete and Partner" DNA
- Even direct competitors frequently partner at the infrastructure level (many rival tools run on Azure).
“It’s the DNA of Microsoft to be competing and partnering with many companies in the industry... many of these AI code generation companies that compete with GitHub Copilot are running their inference on Azure AI Foundry...” (37:54) – Thomas Domke
- Even direct competitors frequently partner at the infrastructure level (many rival tools run on Azure).
- Open Ecosystem: Embracing Rivals as Users
- Meta, AWS, Google, and others are both competitors and GitHub customers/users.
“We see ourselves as part of an ecosystem, as GitHub, hopefully being one of the bigger planets, one of the core planets in Star wars, but there’s a huge system, Mid RIM and Outer Rim...” (40:11) – Thomas Domke
- Meta, AWS, Google, and others are both competitors and GitHub customers/users.
7. Agent Mode: The Coding Team of (AI) Agents
- How Agent Mode Works (52:31)
- Users can assign coding tasks to AI agents, which work in the cloud independently, then submit pull requests for human review—just like another teammate.
“The coding agent is like a new member of your team that can take on certain tasks... you can assign 10 tasks to 10 versions of that coding agent and they can all run in parallel.” (52:37) – Thomas Domke
- Users can assign coding tasks to AI agents, which work in the cloud independently, then submit pull requests for human review—just like another teammate.
- “Vibe Coding” and State of the Art
- Agent instructions are delivered through natural language prompts, issue comments, attached files/images, or even structured internal instructions.
“It’s prompts when you use it within the IDE, although even there you can start with brainstorming cycle first...” (55:57) – Thomas Domke “Whatever your experience was with AI months ago is probably no longer the state of the art now. You shouldn’t really form strong opinions... you should have them loosely held and change your beliefs as AI technology is getting better.” (59:57) – Thomas Domke
- Agent instructions are delivered through natural language prompts, issue comments, attached files/images, or even structured internal instructions.
8. The Future of Coding and SaaS in an AI World
- Empowering, Not Replacing, Developers
- Developers will adapt, automate more, and move “up the abstraction stack”—just as past technologies have done.
“The future of coders or software developers... is bright from my perspective, and software developers will learn very fast, if they haven't already, to adopt AI within their workflows.” (60:54) – Thomas Domke
- Developers will adapt, automate more, and move “up the abstraction stack”—just as past technologies have done.
- What Happens to SaaS?
- Simple, commoditizable applications are likely to be built via prompt, not bought, as SaaS—while valuable, complex SaaS will still justify subscriptions.
“Everything that I can easily replace with a single prompt is not going to have any value... But at the same time I think existing SaaS services will also increase their complexity.” (62:54) – Thomas Domke
- Simple, commoditizable applications are likely to be built via prompt, not bought, as SaaS—while valuable, complex SaaS will still justify subscriptions.
- Pace of Change Is the Innovator’s Dilemma for Everyone
“The back case for any company in our space is missing out on the Next Big Thing... both the bear and bull case are actually very similar, which is can you predict what that looks like and meet developers expectations and have actually the user adoption that then justifies that disruption to your existing business?” (49:24) – Thomas Domke
Notable Quotes & Memorable Moments
-
On AI’s Role Today:
“I think we're underestimating how much you can do with the models that are out today, not the ones that are coming next year or next month. We're only touching the surface of what's possible.” (00:00, 63:20) – Thomas Domke
-
On Developer Empowerment and Democratization:
“AI is going to accelerate the role of the professional software developer and it's also going to enable everyone that wants to become a developer to learn coding.” (01:24, 30:53) – Thomas Domke
-
On the Agent Metaphor:
“The coding agent is like a new member of your team... you can assign 10 tasks to 10 versions of that coding agent and they can all run in parallel.” (00:54, 52:37) – Thomas Domke
-
On Strategic Flexibility:
“There is never going to be one single model that rules them all. The world of software development is just way too broad for this.” (23:27) – Thomas Domke
-
On Open Ecosystems and Collaboration:
“It’s the DNA of Microsoft to be competing and partnering... many of these AI code generation companies that compete with GitHub Copilot are running their inference on Azure AI Foundry and as such they're paying Microsoft...” (37:54) – Thomas Domke
-
On the Future of SaaS:
“Everything that I can easily replace with a single prompt is not going to have any value. It will have the value of that prompt and the inference in the tokens, but that's often a few dollars.” (62:54) – Thomas Domke
-
On What Keeps Leaders Up at Night:
“The bear case for any company in our space is missing out on the Next Big Thing... both the bear and the bull case are actually very similar.” (49:24) – Thomas Domke
Key Timestamps for Major Segments
- Acquisition Rationale & Impact on Microsoft: (03:01–10:26)
- Copilot AI Origins & Early Bets: (11:21–16:09)
- What is Copilot / Explaining to Non-Developers: (16:09–19:49)
- New Era: GitHub Models & Multi-Model Strategy: (21:03–25:31)
- On Fine-Tuning, Tool Use, and Context: (25:31–29:12)
- Market Structure: Players & Segments: (30:15–36:56)
- Microsoft’s Compete + Partner Approach: (37:54–41:29)
- AI Constraints, Open Source and Product Decisions: (41:29–46:44)
- Product Strategy: Infinite Game & Innovators Dilemma: (46:44–52:11)
- Agent Mode Launch & Use Cases: (52:11–57:43)
- Benchmarks and Advice for Trying Agents: (57:23–60:11)
- The Future of Coding and SaaS: (60:11–64:18)
Final Thoughts
This episode distills the excitement and challenges of building, fostering, and competing in the world of AI-powered software development. Domke emphasizes retaining developer-centric values amid rapid technological change, enabling choice and integration for users, and investing ahead in agentic, context-aware future workflows. The takeaways are clear: the future is agentic, collaborative, continually disrupted, and potentially brighter than ever—for those who keep pace.