The Articulator 3000

Inspiration

The theme was Bridge, and we focused on a gap we see all the time: the gap between having an idea and communicating it clearly.

Engineers and builders often understand their ideas internally, but explaining them to teammates, judges, users, or AI coding tools is hard. That gets even harder remotely, where you lose the whiteboard, hand gestures, pointing, and quick sketches that usually help people explain what they mean.

So we built The Articulator 3000: a bridge from messy human expression to clear, shareable communication.

What it does

The Articulator 3000 lets a user explain an idea through:

• Voice
• Typed notes
• Webcam-based air drawing
• Hand gestures and sketches

The app then asks smart clarifying questions and turns the rough explanation into a polished concept page. The final page includes the problem, solution, workflow, visuals, and audience-specific explanations for teammates, investors, or AI coding agents.

How we built it

We built the project as a Next.js app using TypeScript and React. The core experience is a capture studio where users can speak, type, and draw in the air using their webcam.

We used hand tracking to detect the user’s fingertip and pinch gestures, turning the webcam into an interactive canvas. We also built a structured session flow that moves from capture, to clarification, to confirmation, to final generation.

On the backend, we designed API routes for sessions, synthesis, follow-up questions, generation, and sharing. The AI output is validated with schemas so the app can turn messy input into a reliable structured idea model.

Challenges

The hardest part was making the interaction feel natural. Air drawing is fun, but it needs to be stable enough to demo clearly. We had to think carefully about pinch detection, jitter, webcam coordinates, and how to combine sketches with spoken explanations.

Another challenge was scope. There are many things this idea could become, but for the hackathon we focused on one strong journey: capture a rough idea, clarify it, and generate a shareable result.

What we learned

We learned that communication is not just text. A lot of meaning comes from gestures, sketches, timing, and context. We also learned that AI tools are much more useful when they are given structured, clarified intent instead of vague prompts.

The project helped us explore how interfaces can bridge the physical and digital world, especially for remote collaboration.

What's next

Next, we would improve the accuracy of gesture capture, add better visual reconstruction of sketches, and make the generated concept pages more editable and collaborative.

Long term, The Articulator 3000 could become a tool for remote teams, hackathon builders, startup founders, and anyone trying to explain an idea clearly to both people and AI systems.

Built With

• api
• app
• blob
• css
• deployment
• eslint
• generation
• gpt
• hand
• image
• library
• mediapipe
• next.js
• openai
• over
• react
• realtime
• router
• tailwind
• tasks
• testing
• tracking
• transcription
• typescript
• vercel
• vision
• vitest
• webrtc
• zod