The AI-native developer experience | GitHub Universe 2024

Introduction and GitHub's Evolution

• Mario Rodriguez, GitHub's Chief Product Officer, has spent over 20 years working in developer tools and is passionate about providing the best developer tools in the market (29s).
• He believes that software will power everything humanity does from now on, and his goal is to give developers the best tools to build software (54s).
• Rodriguez took the audience down memory lane, showing the GitHub flow from GitHub Universe 2016, which was innovative for its time, and has since expanded to support the entire end-to-end software development life cycle (1m31s).
• Since 2018, GitHub has added features such as GitHub Actions, dependable onemo, GitHub Advance Security, workspaces, and GitHub Copilot, transforming the inner loop of the developer into an AI-native platform (2m5s).
• The AI-native platform enables every developer to build, release, scale, and secure software from an initial idea to production, and is used by over 100 million developers (2m48s).

Building Games with GitHub Spark

• Rodriguez shared a personal anecdote about teaching his daughters how to play poker to learn about probability and luck, but struggled to find a safe and interactive application for them to play (3m56s).
• He showed an example of an unsuitable application he found, which was filled with ads, highlighting the need for better and safer tools for developers and users (4m23s).
• A developer created a fully playable game of Five Card Draw using GitHub Spark, with the prompt "Can you please build a fully playable game of Five Card Draw?" (4m55s)
• The developer also created a game called "Pumpkin Sweeper" using GitHub Spark, a remix of the classic game Minesweeper, but with a pumpkin theme instead of mines, at the request of their daughter (5m16s)
• The developer built a simple math game for their daughter, where she had to answer math questions to get a score, and if she reached a score of 50, she could watch TV (6m17s)
• The math game was built using GitHub Spark, and the code and steps to create it were generated, providing an instructional experience for the developer's daughter (6m45s)
• The developer inserted the generated code into a new file, saved it, and opened it in a browser to play the game (7m22s)
• The developer demonstrated the game of Five Card Draw in the browser, allowing the user to play against the computer (7m47s)

Creating a Personal Game with GitHub Copilot

• A personal game was created using the GitHub Copilot platform to play a variation of Blackjack called "31" with the creator's daughters, allowing them to express their creativity and have fun together (8m19s).
• The game was remixed from a previous version, changing the rules to accommodate the daughters' requests, such as using three cards instead of two and aiming for a total of 31 instead of 21 (9m0s).
• The creator used the Brainstorm feature to generate code for the new game, which not only created the code but also provided a learning experience and encouraged creativity (9m25s).
• The game included features such as calculating hand values and determining the probability of busting, making it a fun and educational experience for the daughters (9m41s).
• The creator's daughters were able to learn about probability and calculation in card games through the game, and the creator could play the game with them remotely when traveling (9m53s).
• The experience highlighted the potential of the GitHub Copilot platform to bring people together and provide a creative outlet, as seen in the creator's personal experience with their daughters (11m2s).

Collaboration at Enterprise Scale with Drum Policies

• A personal idea that felt hard to make happen was to implement drum policies at Enterprise scale, which is now a reality thanks to the team's efforts (13m11s).
• The team created the project on GitHub, breaking down the work into epics, features, and bugs, and using the new sub-issue and issue type features to show relationships between different parts of the work (13m48s).
• The project involved two teams, a lot of features to build, and a nine-week deadline, with tasks added to match how the teams work (13m56s).

GitHub Copilot for Code Reviews

• GitHub's co-pilot is helping speed up code reviews by finding and fixing necessary changes, such as vulnerabilities and stylistic issues (14m57s).
• Co-pilot has reviewed a PR and suggested fixes for a cross-site scripting vulnerability and a stylistic issue, allowing for faster code reviews (15m15s).
• The error message was improved based on co-pilot's suggestion, and the changes were committed back to the project (15m50s).
• Co-pilot's autofix suggestion was applied to fix a simple issue, and the changes were committed (16m15s).
• Developers who use co-pilot autofix fix vulnerabilities seven times faster than those who don't (16m52s).

Improving Security and Portfolio Views

• GitHub is taking steps to improve the security of its actions ecosystem by configuring actions to run as immutable, fully semantically versioned actions, which are distributed as packages and can't be subject to injection attacks, with this feature rolling out to everyone over the next few months, starting with GitHub-authored actions and those from key partners (17m49s).
• Developers can register their interest in having their actions included in the first tranche of immutable actions through a provided link (18m13s).
• Portfolio views are available to track everything being shipped across GitHub, including features like Enterprise rule sets, which allow companies to set flexible and scalable policies across thousands of organizations and repositories in one Enterprise account (18m30s).
• Enterprise rule sets enable companies to target policies based on custom properties of their repositories, allowing developers to create repositories that meet all company security and compliance requirements out of the box (19m2s).

City's Modernization with GitHub Enterprise

• Robert Moore from City is leading development efforts to modernize tools and frameworks, and has aligned his team with GitHub Enterprise as a core service, with over 26,000 developers benefiting from GitHub and Copilot (20m53s).
• City's journey with GitHub involved selecting the platform as a core service and achieving positive outcomes, with Moore set to share the results and initial observations of their experience (21m11s).
• Financial institutions like City are heavily regulated and risk-averse, making it essential to have a secure and compliant development environment (21m19s).
• A new approach to introducing technologies was needed, requiring a trusted ecosystem and a cloud-based offering that could scale and be provided by a reliable partner (21m25s).
• The goal was to provide developers with fully automated, repeatable solutions integrated with CI/CD pipelines, safety nets, guardrails, and controls for compliance (21m41s).
• After an in-depth assessment, GitHub was selected as the partner to achieve these goals (21m56s).
• An internal task force, the Generative AI Task Force, was established to work on the implementation, and GitHub Enterprise was already in production (22m14s).
• The introduction of GitHub Co-Pilot in December 2022 presented an opportunity to bring in a generative AI assistant in the development environment (22m31s).

Introducing GitHub Co-Pilot at City

• The task force realized the need for top-level sponsorship, which was obtained from executive decision-makers, including the CTO and CIOs (23m0s).
• Testing and risk oversight were conducted, including model risk management reviews and three rounds of A/B testing with and without Co-Pilot (23m18s).
• The results showed 10-30% improvement in time to complete tasks, giving confidence to move forward with the Enterprise rollout (23m36s).
• A core team of eight resources was maintained, and a community of practice was built to drive adoption and create best practices, training modules, and a knowledge base (23m47s).
• Super users were identified and empowered to drive adoption, and clinics were held to drive competition and encourage developers to contribute directly (24m3s).
• Gamification was used to encourage participation, and the approach was expanded to every level, region, and business unit (24m40s).
• The impact was consistently measured through surveys, telemetry, productivity metrics, and reporting to ensure the developer experience was improved (25m7s).
• Feedback from developers indicated that they felt more in the flow of their work, with hundreds of examples of positive feedback (25m25s).
• Developers are using GitHub's Co-Pilot to learn new languages and accomplish tasks more efficiently, with junior developers being particularly excited about the tool's ability to help them learn and become more efficient (25m46s).
• Senior developers were initially hesitant but eventually came on board after understanding the benefits of Co-Pilot's assistance in removing mundane and repetitive tasks (26m0s).

GitHub's Vision for Generative AI

• GitHub is aggressively trying to transform its community, developer practices, and experience, and believes that generative AI is essential for its business and developers to attract and maintain new talent (26m52s).
• The company sees Co-Pilot as a must-have integration for its developers to enable them to move at speed and is committed to driving this vision of modernization across developer tools (27m15s).
• City has demonstrated that it is possible to introduce generative AI in a responsible and controlled fashion at scale, with 26,000 developers actively using Co-Pilot (27m47s).
• GitHub is not stopping at Co-Pilot and is working on enabling an extended ecosystem of rich partnerships, including Co-Pilot extensions that will be generally available in early 2025 (28m35s).
• Co-Pilot extensions will allow developers to combine the power of Co-Pilot with other tools they know and love, as well as build their own for their enterprise (28m42s).

GitHub's Partnership with Perplexity

• GitHub is unlocking the ability for developers to stay in the flow as they generate tests, get packaging insights, and open tickets, and has partnered with Perplexity to bring AI-powered search to Co-Pilot (28m59s).
• Perplexity's mission is to cater to the world's curiosity, and its partnership with GitHub aims to accelerate human progress by providing developers with the information they need (29m59s).
• GitHub's partnership with Perplexity allows developers to augment human creativity with the best of the open web, making software development limitless (30m8s).
• A demo showcases the integration of Perplexity and GitHub, enabling developers to stay up-to-date with the latest library updates, get instant answers to questions, and receive API integration assistance directly within the GitHub Co-Pilot platform (30m23s).
• This integration is now available to all GitHub and Perplexity users, and GitHub Universe attendees will receive a free year of Perplexity Pro Plan, which includes advanced PR search, file analysis tools, and cutting-edge AI models (31m4s).
• Perplexity Pro Plan also offers access to AI image generation models, and the company is excited to see what developers will build with Perplexity and GitHub Co-Pilot together (31m32s).

The Future of Development and GitHub's Mission

• The industry is experiencing an incredible time of innovation with the power of new AI LLMs and capabilities, and GitHub's mission is to accelerate human progress by creating 1 billion developers by 2030 (32m7s).
• To achieve this goal, GitHub believes the industry will see a paradigm shift in how we think about code, with developers relying more on specialized agents and natural language to complete tasks and workflows (33m25s).
• The company is committed to keeping innovation, developer choice, and developer experience at the heart of its mission (34m14s).
• GitHub's platform offers a rich extensibility model and powerful network effects, enabling developers to create software quickly and efficiently (34m2s).
• The current landscape presents a wide range of opportunities for creators and innovators to stay in control and shape the future (34m24s).
• The audience is thanked for their presence and participation in exploring this landscape of opportunity (34m28s).
• The future is being shaped collectively, and everyone's involvement is valued in this process (34m35s).