GitHub introduced Project HydraFusion, a research preview delivering frontier-level quality through runtime orchestration. It builds an execution plan that selects models across multiple providers to draft, critique, and revise, or cascades to stronger models for harder tasks. It is available to try in Copilot CLI, with GitHub soliciting user feedback.
GitHub announced HydraFusion is now available in Visual Studio Code and the GitHub Copilot app, expanding beyond Copilot CLI. The expansion was the top request from early user feedback. The release also adds greater transparency into workflow steps and more frequent real-time progress updates.
GitHub announced HydraFusion is now available in Visual Studio Code and the GitHub Copilot app, expanding beyond Copilot CLI. The expansion was the top request from early user feedback. The release also adds greater transparency into workflow steps and more frequent real-time progress updates.
A recorded AI Engineer World's Fair 2026 talk by GitHub Field Copilot Specialist Jose Palafox explains how teams scale custom agents from local machines to organization-wide and CI use. It covers Copilot's internals, the Copilot CLI's built-in agents, the agent builder, marketplace sharing, Orchestrate for cross-repo runs, and GitHub Agentic Workflows for headless CI runs — including a repo with roughly 200 pipeline agents.
A recorded AI Engineer World's Fair 2026 talk by GitHub Field Copilot Specialist Jose Palafox explains how teams scale custom agents from local machines to organization-wide and CI use. It covers Copilot's internals, the Copilot CLI's built-in agents, the agent builder, marketplace sharing, Orchestrate for cross-repo runs, and GitHub Agentic Workflows for headless CI runs — including a repo with roughly 200 pipeline agents.
A developer tested Copilot's preview HydraFusion orchestrator on a medium-difficulty coding task and rated the experience 3/5. HydraFusion auto-selected GPT 5.6-Sol as the primary model and burned 827 tokens, roughly twice the tokens and cost of a manual setup pairing 6.1 Sol with Opus 5.5 for review. The reviewer says they won't adopt it set-and-forget until a model-selection or steering preference ships.
A developer tested Copilot's preview HydraFusion orchestrator on a medium-difficulty coding task and rated the experience 3/5. HydraFusion auto-selected GPT 5.6-Sol as the primary model and burned 827 tokens, roughly twice the tokens and cost of a manual setup pairing 6.1 Sol with Opus 5.5 for review. The reviewer says they won't adopt it set-and-forget until a model-selection or steering preference ships.
Workestrator (also called Crew Forge) is an open-source tool that assembles a team of AI agents from different models (Claude, Codex, GPT, Grok, Gemini) to collaborate on a task. It uses a coordinator agent to dynamically delegate work step by step, supports local CLIs without API keys, and allows agents to use real tools in a scoped workspace. It solves the problem of managing multiple AI subscriptions and switching between them manually.
Workestrator (also called Crew Forge) is an open-source tool that assembles a team of AI agents from different models (Claude, Codex, GPT, Grok, Gemini) to collaborate on a task. It uses a coordinator agent to dynamically delegate work step by step, supports local CLIs without API keys, and allows agents to use real tools in a scoped workspace. It solves the problem of managing multiple AI subscriptions and switching between them manually.
pi-fusion is a tool that combines outputs from multiple LLMs to leverage their diverse strengths and mitigate individual weaknesses. It addresses the problem of single-model limitations by providing a panel of models for tasks like code review, where different models catch different issues. This helps developers and researchers get more comprehensive and reliable AI-generated insights.
pi-fusion is a tool that combines outputs from multiple LLMs to leverage their diverse strengths and mitigate individual weaknesses. It addresses the problem of single-model limitations by providing a panel of models for tasks like code review, where different models catch different issues. This helps developers and researchers get more comprehensive and reliable AI-generated insights.