GitHub Universe 2026: Full Agenda Is Now Live
GitHub Universe 2026 is set for October 28-29 at the Fort Mason Center in San Francisco, and the complete session catalog has just been published. The two-day event features speakers from AMD, Figma, NVIDIA, Coinbase, Anthropic, OpenAI, and GitHub’s own engineering teams, covering everything from practical Copilot workflows to enterprise-scale AI adoption.
Early Bird registration runs through August 19, saving you $300 off the standard pass—and group discounts can be stacked on top. If you want to shape the lineup, you can also vote on which of three community-submitted breakout sessions earns a main-stage slot. Voting closes August 21.
Hands-On Learning and Practical Workflows
A strong thread through this year’s agenda is walking away with something usable: working configurations, reproducible workflows, or a new technique to apply in your own repositories the next day.
Stop prompting, start delegating — Ken Muse and Mickey Gousset (GitHub) demonstrate how to configure Copilot so it can be trusted with complex tasks. The session layers Copilot’s full stack onto a TypeScript application, leaving attendees with both a working project and clarity on which capability fits which job.
Inside GitHub Copilot’s coding harness — Julia Kasper (Microsoft) walks through the evaluation framework the Copilot team uses to test its agent across OpenAI, Claude, Gemini, and future models. Expect details on reproducible benchmarks, thousands of autonomous coding tasks, and LLM-graded assertions that catch regressions early.
GitHub stacked pull requests in practice — Sameen Karim (GitHub) shows how to split changes into smaller, dependent pull requests that move through review efficiently while preserving the full change context. The demo builds a stack via the GitHub CLI, reviews on github.com, and merges each PR as it’s ready.
Community and Organizational Stories
Universe isn’t only about code—it’s also a chance to hear how teams navigate the human side of adoption and collaboration. The agenda includes several sessions focused on community building and organizational change.
Building AI fluency at UPS — Jared Hatfield discusses the gap between developers trying Copilot and developers truly fluent with it. The talk covers how UPS moves teams from awareness to daily use of agents for planning, writing, and shipping, with emphasis on motivation and measurement as much as tooling.
Code is the easy part: Building Home Assistant in the open — Franck Nijhof (Open Home Foundation) explains how the project runs its entire roadmap—not just code—across 20,000 contributors and a dozen GitHub organizations. Ideas, UX, design, and architecture all happen in public across issues, projects, and discussions, including the uncomfortable parts of being wrong and fixing it transparently.
I made my Octolamp think with GitHub Copilot CLI hooks — Beatris Mendez Gandica (Nuevo Foundation) demonstrates using Copilot CLI’s native hooks system to drive physical hardware. Her desk lamp breathes green while the agent works, turns white when idle, and blinks red on errors—all via the WLED API and lifecycle events, with no prompting tricks involved.
What’s Next for AI-Assisted Development
Several talks pull back the curtain on where the industry is heading, from frontier model roadmaps to the operational realities of running agentic systems at scale.
The view from the labs — Cara Phillips (Anthropic), Rohan Varma (OpenAI), and Kate Catlin (GitHub) sit down for an interactive panel on the next two years of AI-assisted development, discussing capabilities only visible from inside the labs that build frontier models.
Fine-grained authorization for hosted MCP servers — Nick Taylor (Pomerium) addresses the problem of hosted MCP servers handing every agent the full breadth of its human’s permissions. He demonstrates an identity-aware proxy that adds per-identity authorization—allowing an agent to open a pull request but not merge it, for instance—without requiring upstream changes.
From writing code to managing agents — Anjuan Simmons (GitHub) shares how the company’s Lifecycle team replaced hand-fixes across 50+ services with an agentic pipeline built on GitHub Actions and Copilot. AI agents classify issues, research codebases, write implementation plans, and open draft PRs autonomously—and the talk includes real lessons on where automation pays off and why some code still gets written by hand.
Choose the Main Stage Session
Three breakout candidates are competing for a main-stage slot, and the winner will also be available on demand after the event. Cast your vote before August 21:
Why sketching in code matters more in the age of AI — Qianqian Ye (Processing Foundation / USC) and John Maeda (GitHub) argue that as AI generates clean code on demand, the human skill that matters is understanding what’s been made. Drawing on teaching creative coding and leading p5.js, the session focuses on building the intuition to shape systems, not just prompt them.
Still open: How AI reshaped open source — Angie Jones (Agentic AI Foundation), David A. Wheeler (The Linux Foundation), and Priya Pahwa (Django Software Foundation) tackle the stress test AI has put on open source: accelerated contributions collide with harder questions about quality, governance, and community at machine scale.
The human side of AI — Brittany Istenes presents how a large U.S. financial services firm reached 94% Copilot adoption and 78% agent-written code—driven not by tooling but by grassroots engagement, storytelling, and innersource practices that made AI something developers actively wanted to use.
Add-Ons and Registration Details
Beyond the main two-day program, attendees can add a Day of Learning pass for $50, which includes a full day of sessions at GitHub HQ plus one GitHub Certification exam voucher. If you already have your agenda mapped out, a certification exam voucher can be purchased standalone for $30 to validate your skills post-event.
