Innovating Developer Tools with AI: Insights from GitHub Next

How do you design AI developer tools that account for LLM hallucinations? Learn how GitHub Next is building controllable co-agents that prioritize structured, task-oriented workflows.

#1about 4 minutes

Investigating the future of software development at GitHub Next

GitHub Next focuses on creating AI-powered prototypes to enhance developer experience, productivity, and happiness.

#2about 6 minutes

Explaining how large language models work and why they hallucinate

Large language models function by predicting the most probable next word, a probabilistic nature that can lead to incorrect or fabricated outputs known as hallucinations.

#3about 5 minutes

How GitHub Copilot was designed to keep developers in flow

The original GitHub Copilot uses inline suggestions and a "ghost text" UI to automate boilerplate code without disrupting a developer's state of flow.

#4about 4 minutes

The impact of ChatGPT and the rise of chat interfaces

ChatGPT's mainstream success created a user expectation for chat-based AI interfaces, which are better suited for planning and exploration than for maintaining coding flow.

#5about 9 minutes

Using Copilot Workspace to turn GitHub issues into code

Copilot Workspace provides a structured workflow that uses AI to brainstorm, generate an implementation plan, and apply code changes directly from a GitHub issue.

#6about 5 minutes

Building and iterating on micro-applications with GitHub Spark

GitHub Spark is a tool and runtime that allows developers to rapidly generate and modify small applications using natural language prompts.

#7about 7 minutes

Designing cooperative and controllable AI agents for developers

Effective AI tools should function as controllable "co-agents" that enhance user capabilities and are designed defensively to handle inevitable failures gracefully.

#8about 4 minutes

Exploring the future of AI beyond simple code generation

The next frontier for AI developer tools includes creating UIs that sync natural language with code, focusing on code understanding, and carefully considering the ethics of AI application.

#9about 1 minute

Book recommendations for prompt engineering and LLM observability

Two recommended books cover the practical skills of prompt engineering for LLMs and the critical process of implementing observability for AI systems.