AI Tinkerers - New York City Hackathon

Restate is building durable infrastructure for AI agents and backends. We're looking to hire our first DevRel Lead to spread the word and bring Restate to developers.

Founding Applied AI Lead - Paxos Health

Build healthcare AI agents in NYC. Apply now.

All Jobs

Final round winners have been announced. View Results

You must be a hackathon organizer or judge to access that function.

Theme

The era of raw AI compute is over; the new bottleneck is human interaction. Explore how powerful models like Claude shatter old UI constraints—sliders, menus, and cards—to forge genuinely novel, playful, and responsive experiences. What does the screen look like now that intelligence is instant?

Detailed Schedule

Judging Criteria

We will evaluate projects based on the following four criteria, focusing on the novelty and execution of the user interface experience enabled by Claude:

Criterion	Description	Scoring (1 - 5)
Working Prototype & Execution	How well the concept is realized into a functional, demonstrable product. Focus is on stability and responsiveness, not just the idea.	1: No working code or major crashes. 2: Runs, but is highly unstable or missing core functionality. 3: Functional, but with noticeable bugs or rough edges. 4: Solidly working prototype with minor polish issues. 5: Flawlessly executing, stable, and responsive application.
Interface Novelty & Playfulness	The degree to which the project explores new UI patterns (breaking convention) and creates an experience that feels fresh, playful, or surprisingly intuitive, moving beyond standard chat windows.	1: Standard chat interface or direct copy of existing patterns. 2: Minor visual tweaks to a standard interface. 3: Introduces one slightly novel interaction or visual element. 4: Successfully implements a genuinely fresh UI pattern (e.g., timeline, canvas, unique keypad) that enhances interaction. 5: Reimagines human-AI interaction with a groundbreaking, playful, and highly usable new interface paradigm.
Theme Alignment: Generative Interfaces	How effectively the project leverages Claude’s power to make interaction itself the innovation, focusing on the interface rather than just the AI’s output quality.	1: Uses Claude only as a backend for standard input/output. 2: The interface is slightly informed by Claude’s capabilities. 3: The interface adapts based on Claude’s state or output in a meaningful way. 4: Interaction is deeply coupled with Claude’s reasoning, leading to adaptive or purpose-driven UIs. 5: The entire experience is defined by a novel interaction loop enabled directly by Claude’s reasoning power (e.g., interaction is the innovation).
Leveraging Claude’s Capabilities	How well the project utilizes the specific strengths of Claude (reasoning, context handling, multimodal potential if applicable) to enable the novel interface, beyond just using a generic LLM.	1: Project could use any basic LLM. 2: Uses Claude, but only for simple text generation. 3: Leverages Claude’s reasoning for multi-step tasks or complex state management. 4: Integrates Claude’s specific strengths (e.g., complex instruction following, long context) to drive the core UI logic. 5: The interface design is impossible or significantly degraded without Claude’s specific, powerful reasoning abilities.

Prizes

Handbook View Handbook

🛠️ Technical Handbook: Generative Interfaces × Claude

Welcome builders! This handbook collects the essential tools, starting points, and inspiration to help you build novel interfaces powered by Claude this weekend.

🚀 Core Tools & Resources

1. Anthropic & Claude API

Claude is the core intelligence for this hackathon. Focus on leveraging its reasoning, context window, and multimodal capabilities (if applicable to your project). Apply for free Anthropic credits here.

Access & Credits: Use the complimentary Anthropic credits provided upon vetting/check-in.
API Documentation: Link to official Anthropic API Docs
Claude Models: Experiment with Claude 3 Opus, Sonnet, and Haiku to balance capability and speed for your UI interactions.
Prompting Guide: Focus on System Prompts to define the behavior of your interface agent, not just the output.

2. Interface & State Management

Since we are focusing on novel UI, robust state management and fast front-end tooling are key.

Redis for State/Vectors: Use Redis for high-speed caching of session state, user preferences, or for vector search if you are integrating RAG into your interface logic.
- Redis Agent Memory Server here
CopilotKit: If you are building an agentic application or want to quickly integrate an LLM into an existing application structure, CopilotKit provides a fantastic framework.
- - GitHub Repo
Generative UI repo
Getting started docs
Getting started with CopilotKit + MPC apps

3. Community Sponsor Tools

Explore how these tools can enhance your experience:

Tavus: While Tavus focuses on human-like video generation, consider how their vision of “instinctive computing” can inspire your interaction models—perhaps your interface needs to feel more human or responsive. see resources here

✨ Project Starters & Interface Ideas

The goal is to break convention. Don’t just build a better chatbot. Ask: “What should the screen look like now that the AI is powerful?”

Temporal Interfaces: Design an interface where Claude manages a timeline of events, allowing users to drag, scrub, or annotate complex sequences (e.g., debugging a process, planning a story).
Constraint-Based Canvas: Build a tool where the user draws/places basic shapes (cards, windows) and Claude automatically structures the content and layout based on those constraints, rather than the user typing out layout instructions.
Keyboard-First Everything: Develop a complex application (like a spreadsheet or design tool) where every action—from selecting a tool to changing a parameter—is initiated via a natural language command typed into a dedicated, persistent command palette, leveraging Claude’s speed.
Adaptive Menus: Create a dynamic navigation system where the menu structure (sliders, keypads, pop-ups) changes contextually based on the user’s current task, inferred by Claude’s understanding of the input stream.

🏆 Judging Criteria Focus

Projects will be judged heavily on:

Novelty: How fresh and unexpected is the UI pattern? (Did you break convention?)
Feel: Does the interaction feel playful, intuitive, or newly possible?
Integration: How effectively is Claude used to drive the interface logic itself, not just the content?

Good luck, and ship something real!

Already registered? Sign in

Quick Actions

Teams Chat Handbook Sponsors

10:00 AM	Welcome Ceremony & Kickoff	Opening remarks by AI Tinkerers, Keynote from
10:15 AM	Sponsor Spotlight: Anthropic	Introduction to Claude, technical guidance, and Q&A with Anthropic engineers.	Sponsors Tauvus, Redis, CopilotKit kick off!
10:45 AM	Official Hacking Begins!	Teams finalize and start building.
11:30 AM	Team Formation Deadline	Crucial Deadline: All participants must be in a finalized team.
12:30 PM	Lunch Break	Catered lunch provided.

7:00 PM	Top five hackers demo live.
8:0 PM	Winners announced
9:00 PM	Event Close / Venue Clear	End of the official hackathon hours. (Final presentations/awards may be scheduled separately or immediately following).

Rank	Prize	Value (Anthropic Credits)
1st Place	Grand Prize	$2,500
2nd Place	Runner-Up	$1,500
3rd Place	Third Place	$1,000