Amelia Wattenberger

Amelia Wattenberger leads product development for Intent, Augment Code’s workspace for coordinating AI coding agents, and is a partner at Sutter Hill Ventures. A designer, engineer, and data-visualization author, she creates interfaces that help people direct increasingly autonomous software without losing sight of the decisions that matter.

From interactive visualization to agent orchestration

Wattenberger began in frontend development and interactive graphics, including work at The Pudding. She wrote Fullstack D3 and Data Visualization and contributed interactive visualizations to the 2019 State of JavaScript survey.

At GitHub Next, she became a principal research engineer exploring AI-powered developer interfaces. Her projects included Code Atlas, which combined language models with structured representations of software; Copilot for Docs; Code Brushes; and GitHub Blocks, an experiment in customizable repository views.

She subsequently joined Adept’s design team, developing interfaces for AI systems that read screens and operate software. Her 2023 AI Engineer talk explored how those systems could assemble personalized comparisons of travel listings, extract practical details, and preserve users’ ability to take action at every level of abstraction.

At Augment Code, she leads Intent’s agent-orchestration workspace. Each task occupies an isolated Git worktree containing its code, plan, terminals, review tools, and agents. A coordinator proposes an implementation plan, specialized agents execute approved work, and a verifier checks the result; external agents including Claude Code, Codex, and OpenCode can also participate.

• Human augmentation through smaller automations: Spreadsheets improve financial judgment by combining many automated calculations within an interface people can control. Wattenberger applies the same principle to agents that summarize information, organize decisions, or implement bounded changes.

• Moving across the ladder of abstraction: Digital maps reveal different details at different scales. Her writing-interface prototypes apply that model to text, transforming paragraphs into summaries and chapters into overviews so writers can inspect pacing and structure without losing access to individual words.

• Workspaces beyond chatbot interfaces: Her critique of chatbot-centered design argues that conversational boxes provide an unnecessarily narrow interface for intelligent systems. Intent instead organizes agents and tools around a persistent, inspectable unit of work.

• Living specifications for coordinated agents: Static specifications drift from implementation as software changes. Wattenberger favors plans that humans approve and agents continually update as they encounter constraints and revise their approach.

• Protecting time for human judgment: Faster development can eliminate opportunities for reflection. She advocates deliberate planning, comparative prototyping, and protected attention so engineers still evaluate whether they are building the right thing.