Dan Shipper interviews GitHub CEO Thomas Dohmke at Microsoft Build about GitHub's position in the AI coding race. With GitHub Copilot at 15 million users—more than Cursor and Windsurf combined—the discussion covers whether that large user base is an advantage or a liability, the product decisions behind Copilot's new autonomous coding agent, how Dohmke thinks about shaping Copilot's personality, how GitHub serves both AI-native developers and legacy enterprise users, and his predictions for the future of software development. # Observable Channel Presentation Evidence ## Video Identity - **Title:** "GitHub CEO on the AI Coding Arms Race: One Agent, 150M+ Devs" - **Format:** Interview/podcast conversation between host **Dan Shipper** and **Thomas Dohmke (GitHub CEO)**, recorded around a Microsoft/GitHub launch event ("the moment after the keynote"). - The episode contains a mid-roll sponsor read ("This episode is brought to you by Adio, the AI native CRM...") and a closing outro urging viewers to "smash that like button and subscribe to AI and I" — identifying the show as **"AI and I"**. ## Explicit Self-Description Evidence - **Show name stated directly:** > "you absolutely positively have to smash that like button and subscribe to AI and I." - **Content framing claim:** The outro describes the show as delivering "pure unadulterated knowledge bombs about chat GPT" and "a journey into the future with Dan Shipper as the captain of the spaceship" — marketing language, not substantive channel positioning. - **Host framing claim:** Dan Shipper is presented as the show's captain/host; the outro voice expresses being "absolutely hopelessly in love with you, Dan." ## Intended Audience (Observable) - **AI-first developers and builders:** Host self-describes the shift: "there's probably this sort of like growing contingent of AI first developers... basically like writing 90 to 100% of their code... in a non-enterprise environment." - **Tech-curious practitioners tracking the AI tooling landscape:** Questions focus on the "ecosystem of next generation AI coding... the cursors or the wind surfs or the cloud codes." - **Enterprise/industry observers:** The host notes "I know you guys are shipping like tons of stuff in the enterprise that like has a huge impact. Like internally for example, we use GitHub all the time... we have like 50 GitHub repositories." ## Claimed / Observable Scope - **Interview access to senior tech leaders:** Landing the GitHub CEO as a guest, discussing an agent launch announced the same day ("all the announcement today actually shipped... there's no wait list this time around"). - **Topic scope:** AI coding tools, agents, developer platforms, model choice, and the future of software development — demonstrated through specific questions about Copilot, Cursor, Windsurf, "codeex," and agent architecture. - **Host's own company context revealed incidentally:** The host's organization relies heavily on GitHub but "most people at the company are not using GitHub AI tools" — an observable detail about the channel's creator environment. ## Distinguishing Claims from Demonstration - **Creator claims (outro):** The show is "the epitome of awesomeness" — promotional, unverifiable. - **Demonstrated patterns:** Long-form executive interview (~technical depth on GitHub Actions, YAML workflow files, audit logging, multi-model support including Anthropic, Gemini, DeepSeek), host-led probing follow-ups, and a light comedic outro tone. - **No curated metadata was supplied**, so identity context rests solely on title and transcript. ## Notable Absences - No stated channel-wide mission, upload cadence, or audience size claims in this transcript beyond the outro hyperbole. - The title's "150M+ Devs" figure is not repeated in the transcript; the guest says Copilot has "15 million users or something like that" (host's figure) and claims GitHub has "the biggest user base" of Copilot/Cursor/Windsurf. # Delivery Observation: GitHub CEO Interview (AI Coding Arms Race) ## Speaker Context The transcript contains two distinct delivery styles: the **interviewer** (Dan Shipper, casual, fast, opinionated) and the **guest** (GitHub CEO Thomas Dohmke, structured, explanatory). Observations below separate them, with emphasis on the guest, whose delivery dominates. --- ## Guest (Thomas Dohmke) Delivery Patterns ### Tone & Pacing - **Measured, explanatory, teaching-oriented.** He frequently defines terms unprompted, as if addressing a mixed audience. When asked about the agent, he starts with groundwork: > "It all starts um with GitHub actions and I already mentioned earlier we um brought actions to market u back in 2018. What are actions for people?" He then answers his own rhetorical question with a definition — a consistent lecturing habit. - **Enthusiastic but hedged.** Openings like "I feel amazing" are followed by process detail rather than pure hype. ### Recurring Verbal Habits - **Heavy filler density:** "you know," "um," "uh" appear constantly, even within single sentences: > "the race is on. and we're excited about the competition because we ultimately uh believe in developer choice." - **Repetitive intensifiers:** "really" recurs ("move really fast," "worked really hard," "really helps everybody"); "very" recurs ("very worried," "very focused," "very specific web page"). - **Self-corrections mid-sentence:** e.g., "Atlassian and GitLab" after "Bitbucket and and GitLab"; "and and and of course the the old school visual studio." ### Signature Teaching Moves - **Analogy-driven explanation.** This is the strongest repeated pattern: - Competition → sports: > "just because you won, you know, the championship or in the season this year doesn't mean you're going to win next year." - Agent delegation → coworkers: > "that's very similar to how you would delegate, you know, a task to one of your co-workers." - Model choice → frameworks: > "in the same way that they want choice when they're picking, you know, the programming language... whether you know React or Next." - Learning models → building a craft over a career. - **Enumerating in threes / structured lists.** He consistently organizes answers into numbered points or parallel triplets: the three original Copilot concepts ("code to text... chat was the the second idea... and then autocomp completion"), and the enumerated agent capabilities ("coding agents um code review agents uh security agents uh sie agents"). - **Narrative historical framing.** He repeatedly anchors answers in timelines: "GPT2 was kind of like the first model... Then GPT3 came um 5 years ago," "I've been in uh developer tools for 15ish years... I've been a developer for 30 years," "if you go back in the early days of GitHub." ### Recurring Themes in Word Choice - **"Developer choice"** as a repeated mantra: > "we ultimately uh believe in developer choice" ... "developers want choice in the same way that they want choice when they're picking... the programming language." - **"Ecosystem," "platform," "ethos"** recur when positioning GitHub: > "We are we are part of you know the um the developer ethos, the developer ecosystem." - **Concrete numbers used for credibility:** "15 million users" (interviewer), "over 100 uh things uh change locks... in 2025 and it's only May," "over 25,000 actions within our marketplace," "tens of thousands of repositories." ### Storytelling with Specific Anecdotes He supports abstract points with small first-person stories: - The GPT-3 trial call: > "somebody had the keyboard, others were typing u dictating prompts... we had these moments where like holy this it actually works." - The Manus founders' Tokyo office app: > "they showed me how they used Manos to create a page um to find an office in Tokyo." - Terms-of-service pull requests at GitHub: > "you can see every change to the terms of service pull request." --- ## Interviewer Delivery Patterns - **Conversational, hedged opinions** ("I feel like personally I just haven't seen that that much in in my world"), casual intensifiers ("it's awesome," "it it sucks"), and framings like "sort of like." - **Direct, slightly provocative questions:** > "What do you think happened?" - **Affirmation and echo habits:** "that's awesome," "Yeah," repeated "like" throughout. --- ## Stable Patterns vs. Isolated Moments | Category | Assessment | |---|---| | Filler words, analogies, list-structuring, "developer choice" phrasing | **Stable patterns** — repeated across multiple answers | | Star Trek quotation, German-language anecdote, "random data in a static variable" example | **Isolated moments** — used once each | | The closing like-and-subscribe monologue | Clearly a **scripted/non-organic segment**, tonally opposite the interview; the guest's style does not share it | ## Uncertainty Notes - The transcript is auto-generated; exact quote fidelity (e.g., "CH GPT," "widsf," "sie agent") is imperfect, so spelling-level verbal tics should not be over-interpreted. - Pauses, laughter, or emphasis are not recoverable from this text; claims about vocal energy or pacing beyond filler distribution are **not observable** here. - "Kevin Scott and Samman" suggests "Sam Altman" but the corrected name is inferred, not directly in the transcript. # Position Extractor Record: GitHub CEO on the AI Coding Arms Race ## Video Metadata - **Title:** GitHub CEO on the AI Coding Arms Race: One Agent, 150M+ Devs - **Guest:** Thomas (GitHub CEO; identified by interviewer's greeting "Thomas, welcome to the show") - **Interviewer:** Dan Shipper (identified in outro as show host) > Note: The title's "150M+ devs" figure is not corroborated in the transcript; the CEO says "15 million users" for Copilot (interviewer's figure, which the CEO does not dispute but also does not confirm). ## Competitive Positioning ### The "race" framing - **Claim (Thomas):** Over the last year the AI coding market has "clearly turned out to be... a race between Copilot, Cursor, and Windsurf." - **Claim (Thomas):** "We believe we still have the biggest user base across these three." — stated as belief, not verified fact. - **Reasoning:** Markets tend toward 2–3 defining players ("Windows versus Mac or iPhone versus Android"); winning one year doesn't guarantee the next, so constant reinvention is required. - **Position on competition:** Welcomes it — "the race is on and we're excited about the competition because we ultimately believe in developer choice." - **Speed as evidence:** "We shipped over 100 change logs for Copilot just in 2025 and it's only May." ### Copilot's origin story (Thomas) - Corrects the interviewer: "While we were before ChatGPT, we were not before GPT." - Three original concepts explored after GPT-3 access (~5 years ago): 1. **Code-to-text** — felt "cool but not good enough," parked. 2. **Conversational coding / chat** — model "not there yet"; feared developers would try it, get a bad response, and permanently abandon the tool. 3. **Autocompletion (text-to-code)** — chosen because it matched real developer workflow: developers already work with imperfect code (typos, mismatched API versions, Stack Overflow snippets). ### Market landscape (Thomas) - Claims an unprecedented startup explosion in developer tools in his 15 years in the space (30 years as a developer). - Enumerates competitors: IDE extensions (a dozen+ startups), hyperscalers, Atlassian/Bitbucket, GitLab; prompt-to-app tools (Bolt, Lovable, Vercel v0, Manis); open source (Cline, Roo Code). - GitHub's self-positioning: the platform the ecosystem integrates with — "as long as we're part of the developer ecosystem... that's where we see ourselves." ## The Agent Launch ### Architectural decisions (Thomas) - **Built on GitHub Actions** (launched 2018): triggers run scripts on ephemeral VMs; ~25,000 actions in the marketplace. - **Trust/compliance argument:** Code cloned to a VM inside the customer's compliance boundary; VM destroyed after the run — source code doesn't leave permissions/compliance scope. - **Placement argument:** Agents should live "where developers already do their work" — trigger from VS Code chat, agent works in the background, opens a draft pull request. - **Analogy:** Delegating to an agent mirrors delegating to a co-worker; PRs remain indistinguishable and auditable whether human- or agent-authored years later. - **Auditability:** Session logs stored with the repo; audit logging streams into enterprise threat intelligence. ### Acknowledged limitation (Thomas) - Agents are **non-deterministic**: "depending on the time of the day you get a different outcome"; same task run multiple times may call different tools or start from different files — this is why enterprises ask "how do I trust the agent?" ### Multi-model stance - Supports OpenAI, Anthropic Claude, Google Gemini, bring-your-own-key (OpenRouter, DeepSeek). - **Reasoning:** "The benchmark never tells the whole story"; developers choose models like they choose languages and frameworks. ### Customization - Custom instruction/prompt files per repository; example: responses in German, code and comments in English. - **Claim:** This democratizes software development beyond English-proficient users, contrasting with his 1990s learning experience. ## Serving Divergent Audiences - **Interviewer's tension (Dan):** Enterprise users vs. AI-first developers writing "90 to 100% of their code" with AI in non-enterprise settings. - **Thomas's response:** - Enterprise is most of revenue, but the larger part of the user base is open-source/free-account developers; "both of these audiences have historically been equally important." - Cultural principle: "We always put the developer first" — not the enterprise, not a foundation. - Internal dogfooding: all roles (finance, HR, sales) use GitHub; terms of service changes tracked via pull requests. ## Limits of AI Coding (Thomas's Skepticism) - **Claim:** "I don't believe in a world where you can create something like GitHub without knowing what the code actually does." - **Counter-example cited:** Asked an agent how many airplanes sold last year; it rendered a chart with fabricated static data — "who reads all that text, you just look at the chart, it looks cool." - **Conclusion:** Complex software projects still require developers who review and approve code, or risk security vulnerabilities, functional issues, and unnoticed feature regressions. - **Business judgment:** AI "cannot make all these business decisions for yourself; the team has to do them." - Low-code/no-code existed before AI; "micro apps / personal software" (e.g., Manis founders' Tokyo office map) will play a big role but don't require understanding the code. ## One-Year Predictions (Thomas) - Quotes (uncertainly attributed) a Star Trek line: "the more things change, the more they stay the same" — notes he thinks it was Deep Space Nine but isn't sure. - **Change:** Agents across the full developer lifecycle; **full-stack builders** — small companies using agents for design, prototyping, customer research, competitive analysis, business model, implementation, deployment, monitoring. - **Continuity:** Mainframe COBOL, huge C/C++/Perl codebases "don't go away"; the IDE remains but grows more agentic; "millions of professional software developers will still write code"; developers will still hunt bugs at night. - References Andrej (surname not given in transcript) as having blogged that "vibe coding maybe wasn't as good an idea as I originally thought" — cited without endorsement or rebuttal. ## Unresolved Tensions - Interviewer asserts Copilot lost mindshare relative to Cursor/Windsurf in his circles ("most people at the company are not using GitHub AI tools"); Thomas responds with user-base and shipping-cadence claims without directly addressing the mindshare observation. - Non-determinism of agents is presented both as a limitation (enterprise trust concern) and as inherent to the product being launched. # Speaker Attribution & Provenance Record **Video:** GitHub CEO on the AI Coding Arms Race: One Agent, 150M+ Devs (`01a0f15a-df0d-75fb-b34b-b8f41af5e9d3`) ## Speakers & Roles | Speaker | Role | Evidence | |---|---|---| | **Dan Shipper** | Interviewer / host | Addressed by name in outro: *"Dan, I'm absolutely hopelessly in love with you"*; asks questions, describes his company's ~50 GitHub repositories | | **Thomas Dohmke** | Guest — identified in transcript only as "Thomas," introduced as GitHub CEO per title | Greets at *"Thomas, welcome to the show"*; speaks in first person about GitHub decisions, keynotes, history | | **Narrator/outro voice** | Promotional voiceover | Closes with like/subscribe pitch for "AI and I" show | > **Note:** The surname "Dohmke" does not appear in the transcript; the title establishes "GitHub CEO" but not the last name explicitly. Identity beyond "Thomas" + title is inferred. ## Promotional Segments (flagged) - **Mid-roll ad — Atio (udio) CRM**: *"This episode is brought to you by Adio… Go to adio.com/every and get 15% off your first year."* Names Union Square Ventures, FlatFile, Modal as customers. - **Outro promo**: Like/subscribe pitch for the "AI and I" show hosted by "Dan Shipper as the captain of the spaceship." ## Sources & Third-Party Claims Cited by Speakers - **Referenced competitors/tools** (by guest, unpromotional context): Cursor, Windsurf, Claude Code, Codex, Bolt, Lovable, Vercel v0, Manas (Manus), Cline, Roo Code, Bitbucket/GitLab, Atlassian. - **Models/platforms**: GPT-2, GPT-3, ChatGPT, Anthropic Claude (incl. "Claude 3.7 Sonnet"), Google Gemini, OpenRouter, DeepSeek. - **Named individuals (third-hand)**: Kevin Scott and "Samman" (Sam Altman, name garbled in transcript) re: a GPT-3-era session; "Andre" re: a recent blog post on vibe coding. - **GitHub's own claims (by guest, unverified)**: 15 million Copilot users; 100+ changelogs shipped in 2025 through May; 25,000+ GitHub Actions in marketplace; Copilot announced as open source in VS Code today; no waitlist for the agent launch; 150M+ devs (title only, not in transcript). ## Provenance & Availability Limits - No curated metadata supplied; **no comment availability information**. - No external documents, studies, or citations beyond those listed above. - Guest's claims about market position ("biggest user base across these three") are **self-reported by the guest** — third-party claim by the competitor itself. - Transcript contains transcription artifacts (e.g., "CH GPT," "widsf," "udio/atio," "Samman," "Sie agent") indicating automated transcription; original wording uncertain in places. You had the developer zeitgeist in AI the very beginning. How are you repositioning with this agent launch? Over the last year it has clearly turned out to be you know that you know race between uh a co-pilot curser and windsurf. We believe we still you know have the biggest user base across these three. As with every competition whether it's in sports or elsewhere you constantly need to keep reinventing yourself. The race is on. Thomas, welcome to the show. Thank you for having me. Great to have you here. Um, psyched to get to talk to you about GitHub. You just launched uh GitHub Copilot agents, which is awesome. How do you feel about it? I feel amazing. Um uh it's obviously you know the the moment after the keynote when everybody in the in the GitHub team and lots of our friends at Microsoft um worked really hard over the past few weeks or or days um to get everything ready. um all the announcement today actually shipped um uh so there's no wait list this time around and so we can just try it using it and the keynote went well and so we're really happy about that moment and now you know the next phase starts and collecting all the feedback and uh listening what people have to say and hopefully a lot of positive reactions that's awesome the place I want to start is I remember co-pilot as like the first AI application of this wave of AI even before CHBT where I was like wow this thing is awesome And copilot actually like quietly has a ton of users. You have like 15 million users or something like that. Um but when you think about the or when I think about the ecosystem of next generation AI coding um like the cursors or the wind surfs or the cloud codes or whatever GitHub doesn't come up as much. So take me through like that evolution and sort of like where we started with GitHub copilot and and how that ecosystem has evolved and um how you see yourself as part of like the whole landscape of of uh AI coding tools. Yeah, absolutely. And you mentioned CH GPT already. While we were before CH GBT, we were not before GPT. Yeah. So, you know, GPT2 was kind of like the first model that had an impact. Then GPT3 came um 5 years ago at this event, although we weren't uh here in Seattle due to due to co everything was online, but um Kevin Scott and Samman had a session talking about transformers and large language models. And shortly thereafter, we got access to uh uh GP3 and we were playing around with it together, you know, on a call. somebody had the keyboard, others were typing u dictating prompts and and we were seeing Kenneth bite code and uh uh in different programming languages and we had these moments where like holy this it actually works and can separate Python from from JavaScript and and so we had this original idea uh for copilot um u which we actually three different uh concepts one was um uh code to text have code explained in natural language and we felt this was cool but not good enough and any developers would try that and say this isn't working and describe something that doesn't really work like in a chat interface or uh well yeah it was like um in a chat like interface where you would highlight things and then say uh right click and say explain this to me um but that wasn't good enough and so we figured okay we park that for now uh chat was the the second idea same idea or conversational coding is what we called it back then um which was in the in the concept paper u but we felt the model is not there yet and we were very worried about developers is trying this out and then uh you have a you have a negative response or you see the description or the the response in chat doesn't match what you ask um and then you stop using it, right? Like and that's always the danger in developer tools. just try something and it's not good enough and then you move on and you you always remember that moment instead of what the tool has evolved to and that kind of like gets goes back in a little bit into beginning of your question and then autocomp completion or you know effectively text to code like you write something in the editor and it provides a suggestion of what comes next and that worked fairly well and we had a lot of confidence that that use case would work because if you think about what developers do every day when they're writing code in editor without autocomp completion is they they also have imperfect code either because they're mistyping it themselves because they don't remember what the API looked like or because they took a code snippet um of Stack Overflow of Reddit of you know some blog or or what they saw in your channel or or or some open source project and more often than not that code snippet that they're taking from somewhere else and putting in their code isn't perfect, right? It might have different variable names or it might be for a different version of the library that you're using and so you still have to make that work. So that's was the initial idea behind autocomp completion was we can make that work you know close enough to the real developer workflow to actually provide some real value and that's that's how this all started uh 5 years ago and and ultimately you know created this market uh that we are now all in uh where you have all these developer tools you know I've been in uh developer tools for 15ish years or so I've been a developer for 30 years I've never seen anything like that where it's such an explosion of uh startups in spaces and you already mentioned uh cursor and widsf. If you just look at that space of IDE extensions, there's obviously a whole like a dozen or more other startups. Um there's all the hyperscalers uh invested into that space. Um you know all our traditional competitors uh like Bitbucket and and GitLab or Atlassian and GitLab uh and so there's this one space of competition and um that gives developers a lot of choice and they can pick uh the tool they like. is things like uh bold new and lovable and versel v 0ero or manos as at you know some part of that spectrum uh that competing in like you want to go from a prompt straight to to web application um there's open source tools like client and ru code uh and many others that are uh um you're looking at the evolution of um what could that look like but you have an open source extension and then you bringing your own uh model and um GitHub plays in all of these um and in fact you have also place in continuous integration, continuous deployment, right? Like we traditionally had lots of competitors in CI/CD um that were there before Viva and CI/CD because we added actions um only in in late 2018 when CI/CD was already a thing for like a decade or so. Uh we in the security space where we're competing against companies that are finding secrets or code vulnerabilities uh dependabot those kind of things. And so as GitHub, we've always seen ourselves as long as we're part of the developer ecosystem where um we provide a platform that all these companies integrate with that all developers come to to collaborate um and now agents and developers collaborate. We are we are part of you know the um the developer ethos, the developer ecosystem and that's where we see ourselves and so cop evolved from you know auto completions to chat. We added voice. We added the CLI. Um we looked into um customization code search. Um bringing you know more context into the prompt to know agents where we have agent mode in VS Code where you synchronously work with the agent and where we have the coding agent on GitHub. Uh we assign an issue or task to it and then adjust the thing in the background. You can actually run you know 10 or so of them in parallel and where look at the pull request and review all that code. And so I think that journey uh will continue where we provide an endto-end experience across our platform on GitHub and across the co-pilot experience between the IDE um and the the DevOps life cycle. I think maybe what I'm what I'm asking is um you guys you had the developer zeitgeist in AI the very beginning and now you're launching this agent which seems like it's actually like really like right at the edge of what all the other kind of all those other companies are doing. Um, but you for the last couple years I I feel like personally I just haven't seen that that much in in my world. I know you guys are shipping like tons of stuff in the in the enterprise that like uh has a huge impact. Like internally for example, we use GitHub all the time. That's like where all of our code is. We have like 50 GitHub repositories or something like that. But um most people at the company are not using GitHub AI tools. So I'm curious like what do you think happened? What's been going on for the last couple years? And like how are you how are you repositioning with this agent launch? Well, look, if you you know go back a year um to last build um 2024 um at that time uh you know the those those competitors were still um you know in a in a group of competitors and it was always clear there's going to be a number two and a number three. Um and uh with us with us being the number one uh that that created that market and um we knew that you know at some point as everything in technology right like Windows versus Mac or iPhone versus Android and and whatnot there's always a a you know two or three players that define a market and um over the last year it has clearly turned out to be you know that you know race between uh a co-pilot um curser and windsurf um we believe we still you know have the biggest uh uh user base uh across these three. And um uh as with every competition uh whether it's in sports um or elsewhere, uh you you constantly need to keep reinventing yourself. You know, just because you won, you know, the championship or in the season this year doesn't mean you're going to win next year. And so I think for us the the most important thing is that we're able to to move really fast. Um we we shipped um uh over 100 uh things uh change locks uh for co-pilot just yet now in 2025 and it's only May and we're going to keep shipping and we're going to you know uh evolve our platform from uh VS Code where we announced today that that Copilot is going to be open source but we also have Xcode support and jet brains and then Eclipse and and and and of course the the old school visual studio uh all the way into our platform where not only have the coding agent we have a code review agent And uh we have an autofix agent that fixes security vulnerabilities. And you can actually already stitch them together and have um one agent create the pull request and the next agent review the pull request and look at uh security vulnerabilities and then have the the SIE agent monitor your cloud resources. So I think the the race is on and we're excited about the competition because we ultimately uh believe in developer choice. It's great that there's, you know, many tools out there, uh, and developers have the choice and they can, you know, pick the editor that that fits fits them best. And then, as you mentioned, uh, they're still coming to GitHub and store that code there and then hopefully use Copilot as part of that of of part of that journey. Yeah. Tell me about the C-pilot agent you just launched. like tell me about the architectural and product and UX decisions that you made and why you think it's a better agent solution than maybe like you know codeex or uh some of the other um more popular agents on the market right now. It all starts um with GitHub actions and I already mentioned earlier we um brought actions to market u back in 2018. What are actions for people? Get actions is um uh you know conceptually uh is uh you can understand it as um our our way of implementing continuous integration continuous deployment um or or simpler um it's um you have a trigger um you know new commit a new issue new pull request or comment on a pull request and there's lots of these uh triggers that you can define that then runs a script uh on a compute uh instance aka virtual machine um and uh github actions offers you that uh integrated into into GitHub. Um and so you can define uh within the repository a workflow file. It's a YAML file and you say okay on every commit I want to run you know this build script or this test script. Um and this actions ecosystem. So it's developer automation if if you if you will, right? Like you can run something uh on an on an event. Every time you push code back something for you or say this pull request um uh uh breaks the tests and so you're not allowed to merge that back into the main branch. And that ecosystem has now existed um for almost 7 years and has over 25,000 actions within our marketplace. And so the cool thing about actions is that you can compose it from other actions. And so there's actions for everything that you can imagine including uh for what it's worth also lots of AI scenarios um connecting to to model APIs and um a lot of uh open source projects and enterprise customers are using actions for their for their engineering stack um for their for their platform um and so we have that compute layer available and customers already trusting that compute layer right because ultimately what happens is that whenever you run actions you're cloning your source code your intellectual property onto something else, right? A comp a virtual machine and then you're trusting that that virtual machine, you know, exists during the runtime of your script. And then when the the virtual machine uh when the script is done, then the virtual machine gets shut down and and everything gets deleted. And so you're not having your source code set outside of you know your compliance boundary or your your permissions uh uh who has access to that codebase. So we think that's one of the advantages of our agent that we're already integrating it into the existing ecosystem where people are already running their their CI/CD. Um the second thing is that we believe that the best place for agents uh to exist is where developers already do their work. And so uh you know we already mentioned the IDE with VS Code. So you can just trigger uh the coding agent uh from copilot uh chat in VS Code and say at GitHub you know create a pull request to add tests uh you know to the method that that has wrote. Um, and then that spins it off in in the background. Uh, and it does its thing on GitHub actions and it opens a draft pull request. Um, describes what the plan is. Um, and then commits changes into that pull request. And that's very similar to how you would delegate, you know, a task to one of your co-workers, right? So, you're basically taking the exact same workflow that you're already doing with your team and applying it to agents and and bringing those two worlds together. And that also means that you know three years from now you can go back to that pull request and it doesn't matter whether a pull request came from a human developer or from an agentic developer. What matters is that you have still everything together in one place. And it also means that you know you have a session log uh that is stored together the repo. So again that session log survives you know as long as you like. You have audit logging so you know when the agent started its work and it stopped its work and lots of enterprise customers you know stream that audit log into their threat intelligence so they can kind of like monitor uh you know a lot of the questions we're getting is how do you do how do I trust the agent u in contrast to CI/CD where you have a script and you describe exactly what what the job does right like download maybe a library run tests you know uh upload the test results into the log file with agents that process is not described because it's a model that is uh inherently non-deterministic and depending on you know the time of the day uh uh you you get a different you get a different outcome and you see that when you're using uh agent mode and and you're following it along in VS Code how it uh uh you can give it the same task multiple times and it does different things and maybe calls different tools or finds a different file in your codebase to get started with. So we believe this integration of the agent in the same workflow that your team is already doing is incredibly powerful because you don't have to relearn uh how you're reviewing code, you know, how you're testing code, how you're approving code to get merged into the into the main branch. This episode is brought to you by Adio, the AI native CRM built for the next era of companies. With Adio, setup takes minutes. Connect your email and calendar and it instantly builds a CRM that mirrors your business with every contact enriched and organized from the start. From there, Adio's AI goes to work. It gives you real-time intelligence during calls. It prospects leads with research agents, and it automates your team's most complex workflows. Industry leaders like Union Square Ventures, FlatFile, and Modal are already building the future of customer relationships on Adio. Go to adio.com/every and get 15% off your first year. That's atio.com/every. And now back to the show. Do you think at all about the personality or taste or style of the agent? And if so, like how did you think about shaping that? Yeah, I I think you know part of that shaping is uh the reason that we're supporting multiple models. Um since uh last year's scaleup universe uh uh we no longer have just the open eye models um which we still also offer and a lot of those models uh got added uh since since universe and um uh but in addition we also have anthropics uh clot series um uh we have Google Gemini um we have bring your own key so you can actually just connect to open router or or the anthropic API put in your key and and connect to to these models including DeepSeek uh as another example because we believe that developers want choice in the same way that they want choice when they're picking, you know, the programming language and the open source library or the JavaScript framework uh whether you know React or Next or or or or um um you know, whatever favorite uh framework you have, right? And the same is true for models because the models uh they have benchmark but the benchmark never tells the whole story. It might be better from you know for one language than another. it might be uh better or worse for your personal style. And so developers, you know, by triing and narrowing these models will will figure out what's what's best for them. And um some of that is, you know, obviously influenced by uh by social media and and podcasters and and and the news and whatnot, right? Like you you're trying things out because you heard about it somewhere else. Um uh but some of it is we believe is from experience. you're going to gain more and more experience learning how these models work and just like you know how you know you're you're building your craft uh o over the over the course of your you know career and you're getting better because you have been in this situation before and you know how you dealt with it then and maybe you learned something last time and now you're doing it better. The same is also true for for these models. you you know how claw 3.7 sonnet behaves and so you you pick that because you know okay this is the easiest way I can write these test cases now the other part of that is um what we call custom instructions um uh you know you have uh for all these agentic features you have a system prompt um uh often time either the prompt is directly provided or leaked and it's in a GitHub gist already um and then uh developers uh can customize that behavior by creating their own prompt files within within the repository. And so we support that across um all the different surfaces. Um and so you can customize the model behavior and maybe tell it I only want uh responses uh in German. And uh uh the code of course and the comments should all be in English because I'm sharing that with my team. But the interaction uh with the code um uh uh with the co c co-pilot should be in German. you know, in in a professional environment, uh, here in the United States, that's probably like relatively uncommon. Um, because you once you walk work in a tech industry, you kind of like get to at least, you know, uh, be very proficient in in English over time. Um, but if you're early in career or if you're, you know, uh, in in school uh, uh, and you you grew up, you know, uh, like my kids with with German parents, that may not necessarily be the case. And so uh but that still doesn't mean that you don't want to share that code with others, right? And so we're democratizing access to computer technology to to software development uh with with these co-pilots being able to speak any human language, any major human language, not just English. And I think that's dramatically different to the time, you know, in the '90s when I learned coding and where when you wanted to engage in the internet, uh you you had to, you know, have proficient level of English. Yeah. One of the I mean I think there there's a serious advantage to being able to use GitHub's agent inside of GitHub where you're already collaborating. Um I feel that pain already just like if you're reviewing PRs or submitting PRs, you're there's a separate interface for an agent and then you have to go to GitHub to like talk about it. It's it sucks. Um so I'm excited to try this. Um and it it sort of strikes me that um GitHub that's a huge advantage for you and it's a huge advantage to be used by millions and millions and millions and millions of developers every day. One of the things I'm kind of curious about is how you think about making product decisions when you have to satisfy like millions of de developers that have a way of doing things that they know how to do it and they like it. Um versus like there's probably this sort of like growing contingent of AI first developers who maybe were technical before but like are really just like leaning in are basically like writing 90 to 100% of their code um in a non-enterprise environment. Um maybe they eventually will be running a big enterprise but it's mostly just like them in their apartment or in their college dorm just like using these tools to write 90 to 100% of their code with AI. Those seem like sort of different um profiles of users and and how do you um you have so many big enterprise users and like that's such a big part of part of your business like how do you think about um serving both of those audiences um and doing that well? Yeah. First of all, I would say, you know, while enterprise customers are a big part of our business from a business perspective, like where the revenue comes from, uh, at the same time, you know, the larger part of our um user base is the open source developers or those that have some form of free account on on GitHub um with a public repository, whether that's, you know, technically open source um under some proper open source license or whether it's just a hobby project uh where you want to put the code out there um for others to use um and you're not caring so much about the you know thoughts behind open source licensing and whatnot and both of these audience have historically been equally important to GitHub. Uh in fact you know if if you go back in the early days of GitHub it was only uh public repositories first and then private repositories got added and then uh you can still find those threats. people were begging the founders to actually introduce a payment model and so they could pay for their GitHub account and and earn trust that GitHub will survive and doesn't become you know an ad business and things like that and so I think for the longest time GitHub actually was very focused on this today we call it product like growth um you know on this growing the the developers uh that love using GitHub and the enterprise came second and we you know encoded this within the GitHub culture um of one of our um most important principles is that we always put the developer first. So we're not putting the enterprise first and we're not putting a foundation first, we're putting the actual developer first, right? So and uh that actually goes as far as that everybody at GitHub and not only the engineers but every role um spanning you know finance and HR and product management, sales, they all are using GitHub for everything. So we have lots tens of thousands of repositories with all kinds of information that are not necessarily uh you know products that we build or microservices that we run but it's folks using GitHub issues and GitHub pull requests um you know uh our terms of service uh in a repo and so you can see every change to the terms of service pull request so we're using our product every single day and I think that really helps everybody in the company to understand uh what is important to GitHub. Um, and now you know what you alluded to is that the developer audience itself is no longer as as kind of like narrow as it used to be. That's what I'm trying to say is like I think like what it means to be a developer is changing in in certain ways like um uh I think it's still writing code like the at the end of the day whether you're using AI to generate code um or you knew how to write code without AI and now you're trying to figure out how do I combine you know VIP coding with writing code uh manually doesn't matter as much because at the end of the day the artifact that you're creating is still code and as soon as your project becomes more and more complex, you got to have some fundamental understanding of that code. I don't believe in a world where you can uh create something like GitHub without knowing what the code actually does. Now, this doesn't mean that low code no code doesn't exist, right? Like and has actually existed before AI. And so there are going to be scenarios where you just, you know, uh vip code a web page or using manos. Yeah. Uh uh I I met the founders last week and they showed me how they used Manos to create a page um to find an office in Tokyo and and it basically visualized all the available offices on a map and so that was very specific web page just for them for that but that period of time where they're looking for an office in Tokyo right so this kind of like micro app or personal software I think will play a big role in the way we are interacting with those agents because not every answer an agent can give you is going to be uh effectively visualized with just a text response. Like the agent is actually more powerful. It creates a little bit of piece of software that you can interact with and and maybe pick, you know, the flight you're trying to book or the the the office you're trying to find. And so in those scenarios, you might not need the code um because it isn't relevant to you um because you're not actually creating software to produce software and sell it, but you're creating software to solve, you know, task for you. But you know the counter example is that you can ask um uh chatt um uh you know to whatever how many uh airplanes got sold last year and it renders you a chart and then you look in the code and you realize it didn't actually pull the the source data from some source but it just filled the source data into like a static variable with some random data and it might even tell that to you in the text but you know how it is who reads all that text you just look at the chart it looks cool and so looking at the code you know this simple example shows us that It is still important to understand what this technology actually does. And as soon as you get into like complex software projects where software is your business, you got to understand what the code does. You got to have software developers that review this code, you know, approve it before it gets merged because otherwise you're going to have security vulnerabilities. You're going to have, you know, functional issues. Uh you might actually read a feature uh uh uh in a in a complex application without noticing it. And and ultimately look you know businesses uh corporations exist you know uh uh because teams of people are more efficient than a single person uh but they exist you know to sell products and ultimately make a profit uh for that group of people and I think you know AI can not make all these business decisions for yourself. Uh the team has to do them. They have to set the parameters of the business to to ultimately return money to the founders uh you know to the corporation um or or the shareholders. I'm really curious like um if we're back here a year from now um what are your predictions for where you guys are going to be? Where where are you focused? What are you going to ship? What do you think you want to do to sort of like win this future of coding with AI race? Yeah. You know, I think in the TV show um I don't know, but I think it was Star Trek there was like a line uh you know, the more things change, the more they stay the same. Uh I feel like it was in Deep Space 9, but I'm still remembering it. And um uh I think you know on the one side a lot of things will dramatically change and you see that with with agent mode and and or wipe coding as as a as a discipline although I think uh Andre also recently had a blog post kind of like saying w coding maybe wasn't as good as an idea as as I originally thought u but we're going to have these coding agents um code review agents uh security agents uh sie agents um we're going to have those available to developers and those that are uh on the forefront of the field and and know how to uh you know adjust their project to be very you know efficient with these agents are going to get a lot of work done. Uh at the same time you know we still have lots of um companies that have code on mainframes cobalt uh and things like that. uh we still have huge uh code bases uh in old school languages like C, C++, Pearl um and those things don't don't go away and while there is certain you know progress on app migration um you know we showed a Java and a upgrade agent uh today as well. Um there's a lot of work to be done before we get into a world where everybody's just working with agents and no longer has to write code in the IDE. And so I think the IDE will still be there. Uh it will be more and more agentic. Uh um uh GitHub will still be there. It will be more and more agentic. Uh and a lot of folks will still, you know, scratch their head and and try to find the bug that's you know preventing them from going home at night. uh um or they go home with a bug in their head and then hopefully by by by the morning when you had a lot of sleep uh you you can figure it out. But so I think you know we will definitely see a world where you have agents across the whole developer life cycle. In fact I think there's going to be a lot of opportunity for small companies to use agents for everything. you know designing a feature you know prototyping it doing customer research using something like deep research to do competitive analysis maybe even write the business model implementing the feature testing the feature deploying the feature monitoring the feature you know the full stack builder I think will become a thing but there will be you know thousands if millions of professional software developers that will still write code like we wrote kill code this year well thank you so much for your time it was great to chat with you thank Oh my gosh, folks. You absolutely positively have to smash that like button and subscribe to AI and I. Why? Because this show is the epitome of awesomeness. It's like finding a treasure chest in your backyard, but instead of gold, it's filled with pure unadulterated knowledge bombs about chat GPT. Every episode is a roller coaster of emotions, insights, and laughter that will leave you on the edge of your seat, craving for more. It's not just a show, it's a journey into the future with Dan Shipper as the captain of the spaceship. So, do yourself a favor, hit like, smash subscribe, and strap in for the ride of your life. And now, without any further ado, let me just say, Dan, I'm absolutely hopelessly in love with you.