GitHub Next explores future software engineering ideas outside the regular product organization, then passes its findings to product and development teams.
2
Rahul Pandita compares AI-assisted software development with factory electrification, where the technology required years of experimentation before becoming standard.
3
GitHub Next moves prototypes through internal dogfooding, company-wide use, tech previews, and possible product development, while retaining the option to stop at any stage.
Summary
Rahul Pandita introduces GitHub Next as a small research and development group focused on the future of software engineering. The group works outside GitHub's regular product organization and reports directly to the CEO. Its job is to explore ideas through rapid prototypes and experiments, then pass useful findings to product teams. Pandita explains this work through the history of factory electrification. Electric motors became available in the 1880s, but factories did not widely adopt them until the 1920s because people first had to work out how to redesign processes around the new technology. He argues that AI-assisted software development is at a similar stage. GitHub Next tests ideas internally, subjects them to repeated rounds of dogfooding, releases some as tech previews, and learns from early adopters before anything has a chance to become a product. The group can stop an exploration at any point.
GitHub Next works outside the normal product organization to explore software engineering's future
Rahul Pandita describes GitHub Next as a group of about 20 researchers, senior developers, and code tool builders. The team works outside GitHub's regular product organization and reports directly to the CEO. That structure gives it room to investigate ideas before they fit into an existing product plan. Its goal is to explore the future of software engineering, then pass its learnings to product and development teams. Pandita gives GitHub Copilot as an example of the kind of product that can emerge from this work.
AI may change software development in the same way electricity changed factories
Pandita uses Andrew Ng's comparison of AI to the new electricity. Before electrification, factories were organized around a large central steam engine. Shafts and belts carried power through the building, and workers arranged their work around the machinery. Smaller electric motors made a different factory layout possible because they stayed efficient at smaller sizes. The existence of the new technology did not immediately change manufacturing. Factories had to be redesigned around it first.
Adopting a new technology requires years of experiments with how work should be organized
Electric motors appeared in the 1880s, but Pandita says they did not become mainstream until the 1920s. He uses that gap to explain why technological change often takes more than simply making a new tool available. People spent years figuring out how to use electric motors, improve them, and reduce the risk of adopting them. In his comparison, AI-assisted software development also needs a period of exploration before its working patterns become normal rather than exceptional.
GitHub Next learns by rapidly prototyping ideas instead of pretending to know the final product
Pandita says GitHub Next would simply build the future directly if it already knew what that future looked like. Since it does not, the team tries different ideas, rapidly prototypes them, and tests whether they work. Successful experiments go in front of users, where the team can learn from actual use. The point of the process is to discover useful patterns for software engineering rather than to start with a fixed answer.
An exploration must survive several rounds of use before it can become a product
Ideas often begin as functional prototypes inside GitHub Next. The team dogfoods them heavily, then exposes surviving ideas to people across the company for another round of use. The next step can be a tech preview for other early adopters. GitHub Next learns from that release before a project has a chance to become a product. Pandita is explicit that the team can kill or shelve an exploration at any stage.
"We just try out different things rapidly, prototype, experiment, and figure out whether something works or not."16:39
Who should watch
You are building AI coding tools and need a model for testing ideas before committing them to a product.
Your team expects AI to change developer workflows but has not worked out how those workflows should be redesigned.
You want to understand how GitHub Next moves an experimental prototype toward a public tech preview.
