From Idea to Demo in Two Days: Inside Superhuman’s 2025 Global Hackathon (opens in new tab)

Superhuman’s 2025 hackathon brought together nearly 500 employees to prototype innovative product features by leveraging cutting-edge AI coding tools like Claude Code and Cursor. By integrating AI-driven agents and keyboard-centric workflows, teams demonstrated how rapid experimentation can bridge functional gaps across mail, documentation, and collaboration platforms. The event highlighted a significant shift toward "vibe-coding" and accessible development, where cross-functional teams and non-engineers could ship functional MVPs in just 48 hours.

Superhuman Command Everywhere (SCE)

• This project extends the Superhuman Mail Command Center to the browser, allowing users to trigger Grammarly features, set reminders, and snooze items from any web page.
• The tool enables keyboard-only navigation for AI agents; for example, users navigate Grammarly’s Proofreader cards using "J" and "K" and accept or dismiss suggestions with "E" and "D."
• Developers used AI tools to quickly interpret an unfamiliar codebase, allowing engineers without frontend expertise to "vibe-code" a working MVP within a few hours.

Whiteboarding in Coda

• This feature introduces a native canvas within Coda documents where users can draw freely, add shapes, and import images for brainstorming and diagramming.
• The prototype includes an AI diagramming tool that generates editable visual versions of diagrams based on plain-text descriptions.
• Built by a solo team member with no formal coding background, the project utilized Claude Code and Cursor to focus on UX refinement and smooth interactions rather than just technical functionality.

Superhuman Listening

• This system centralizes fragmented customer feedback from tools like Gong, Salesforce, and Zendesk into a single, queryable source of truth.
• By linking unstructured data to product roadmaps in Coda, the tool helps sales engineers and product managers determine if specific customer feedback is already being addressed.
• Technical challenges included using LLM APIs to extract urgency and sentiment, though the team noted the difficulty of filtering "noise" from high-volume sources like Zendesk tickets.

Inclusive Language Agent

• Developed by a team of linguists, this agent identifies non-inclusive phrasing or unconscious bias in professional writing.
• The goal is to provide real-time suggestions that improve workplace culture and customer trust by making word choices more inclusive and intentional.

The results of this hackathon suggest that AI-assisted development tools are significantly lowering the barrier to entry for complex product builds. For organizations aiming to accelerate innovation, encouraging "maker" identities across all departments and utilizing AI to bridge technical skill gaps can surface high-value solutions that traditional product cycles might miss.