Ace — GitHub's Bet That Coding Agents Need a Multiplayer Mode
Maggie Appleton's GitHub Next talk hit Hacker News today. The thesis is uncomfortable for anyone selling solo-developer coding agents: implementation is no longer the bottleneck, alignment is.
Ace is GitHub Next's research prototype, opening up a technical preview to a few thousand people. It is a real-time multiplayer coding agent workspace. Picture Slack chat plus GitHub plus Copilot plus a fleet of cloud microVMs in one room. Multiple humans share the conversation, share the same sandboxed environment, share the same agent context. Anyone can prompt the agent. Designers and PMs sit in the same channel as engineers. Live previews update for everyone in the session. VS Code editing is collaborative in real time.
Appleton's argument is that the current agentic dev experience compresses implementation from weeks to minutes, but leaves planning at zero and review as the only alignment checkpoint. By the time someone reviews the PR the agent already wrote ten more. Her line: implementation is rapidly becoming a solved problem, the hard question is no longer how to build it, it is should we build it. Ace makes planning continuous instead of a discrete step.
This is a different bet from where most of the agent stack is going. Cursor, Codex, Claude Code, Windsurf are all optimizing the single-player loop. Ace says the loop itself is wrong, because software was never single-player. If GitHub ships this, it pulls Copilot's center of gravity from autocomplete to team coordination, which is a fight Microsoft can win. The video and signup are at maggieappleton.com/zero-alignment. Watch this one even if it stays research, because the thesis will reshape how everyone thinks about agent UX in 2026.
← Back to all articles
Ace is GitHub Next's research prototype, opening up a technical preview to a few thousand people. It is a real-time multiplayer coding agent workspace. Picture Slack chat plus GitHub plus Copilot plus a fleet of cloud microVMs in one room. Multiple humans share the conversation, share the same sandboxed environment, share the same agent context. Anyone can prompt the agent. Designers and PMs sit in the same channel as engineers. Live previews update for everyone in the session. VS Code editing is collaborative in real time.
Appleton's argument is that the current agentic dev experience compresses implementation from weeks to minutes, but leaves planning at zero and review as the only alignment checkpoint. By the time someone reviews the PR the agent already wrote ten more. Her line: implementation is rapidly becoming a solved problem, the hard question is no longer how to build it, it is should we build it. Ace makes planning continuous instead of a discrete step.
This is a different bet from where most of the agent stack is going. Cursor, Codex, Claude Code, Windsurf are all optimizing the single-player loop. Ace says the loop itself is wrong, because software was never single-player. If GitHub ships this, it pulls Copilot's center of gravity from autocomplete to team coordination, which is a fight Microsoft can win. The video and signup are at maggieappleton.com/zero-alignment. Watch this one even if it stays research, because the thesis will reshape how everyone thinks about agent UX in 2026.
Comments