GitHub's research team, featuring Eddie Aftandilian, Idan Gazit, Russell Horton, and Terkel Gjervig, present their experimental work on Continuous AI and agentic workflows for software teams.

Exploring Continuous AI with GitHub Next: Agentic Workflows for Developers

Speakers: Eddie Aftandilian (Principal Researcher), Idan Gazit (Head of GitHub Next), Russell Horton (Staff Researcher), Terkel Gjervig (Staff Research Engineer) — GitHub

Introduction

Continuous AI represents a shift in software development workflows by leveraging AI agents that perform ongoing, automated tasks in GitHub environments. Unlike traditional CI, these agentic systems use LLMs to handle complex assignments, such as code localization, accessibility improvements, playtesting, performance optimization, and issue triage.

What is Continuous AI?

• Definition: AI agents integrated into development workflows, performing operations autonomously and continuously.
• Difference from CI: Goes beyond rule-based automation, allowing for tasks that require judgement and context.
• Use Cases: Code review assistance, localization, accessibility, test coverage, performance checks, and triage.

Demos and Examples

• Agentic Workflows: Open-source tool for building custom agent automations in GitHub Actions.
• Showcases:
  • Localization and accessibility support via AI automation.
  • Agentic playtesting and coverage analysis.
  • Automated performance optimization and intelligent issue triage.
• Tool Mention: Pelli used for workflow demonstrations.

Safety and Best Practices

• Safety in Continuous AI: How to ensure agentic tasks remain predictable and secure within automation pipelines.
• Getting Started: Guidance on integrating agentic workflows using GitHub's open source tools.

Resources & Further Learning

• Watch more from GitHub Universe: GitHub Universe playlist
• Agentic Workflows repo: (search GitHub for code samples and libraries)
• Connect with GitHub Next:
  • YouTube
  • Blog
  • Insider newsletter

Key Takeaways

• Continuous AI brings judgement-driven automation to developer workflows.
• Open-source agentic tools allow teams to tailor and scale AI-powered CI.
• Practical demos highlight real use cases like test coverage and performance optimization.
• Developers are encouraged to experiment and prepare for agentic practices as part of their future software strategies.

For more practical engineering content from GitHub Next, follow the video, blog, and newsletter channels.