Inspiration
Every fintech app on Earth looks the same: a flat dashboard, a row of numbers, a stack of tabs. You don't feel your money — you read about it.
CORTEX started from a simple, stubborn question: what if you didn't have to learn an app's interface at all? What if you could just say what you wanted, and the app reshaped itself around you? Not a smarter dashboard — the disappearance of the dashboard entirely.
We built CORTEX to prove that idea in the most demanding way possible: apply it to something as mundane and universally understood as checking your spending, and see if immersive 3D and intent-driven navigation could make even that feel alive.
What it does
CORTEX rebuilds a fintech app as a first-person 3D world. There is no menu bar, no tab navigation, no dashboard grid.
- You enter through a cinematic hyperspace sequence and arrive standing inside the Void — a vast, open space where financial features exist as enormous, distant landmarks, visible the way a castle would be visible through mist: far away, but unmistakably real and huge.
- You walk or fly toward what you want, or simply speak or type your intent — "show me my spending" — and the world carries you there directly.
- Each feature materializes using an Orbit model: a central hub representing a category (like Accounts or Spending), encircled by rotating rings carrying smaller orbiting spheres — each one a specific sub-feature you can call up, fly to, or leave in orbit for later.
- A small ambient companion, the Wisp — a soft, coiling ribbon of light — drifts beside you throughout, offering a gentle nudge if you hesitate, and receding the moment you're confidently in flow.
- When you're done, you leave through a mirrored exit sequence, back out of the world entirely.
The vision extends further than what's live in this build: a five-level Deep staircase for more complex tasks, and a hidden, invitation-only Core layer for the app's most powerful features — reached through a deliberately dramatic "the world breaks and a new one takes root" transition. Those layers are fully designed (see "What's next") but intentionally not the primary build for this submission — see "Challenges," below, for why.
How we built it
- Vite + React + React Three Fiber for the 3D scene graph and rendering
- @react-three/drei for particle systems, text-in-3D, and camera helpers
- @react-three/postprocessing for the bloom/glow that defines the whole visual language
- GSAP for every eased camera movement — the entrance warp, the "boost" toward a landmark, the Orbit hub fill — so nothing ever snaps or cuts abruptly
- Web Speech API for voice input, with a plain-text fallback, both feeding a lightweight keyword-to-intent matcher that routes the user to the right feature — no heavy backend or ML model required to make the world feel responsive
- Vercel, connected directly to GitHub, for continuous deployment — every push rebuilds and redeploys automatically
All financial data in this build is realistic mock data — there's no real account integration, and intentionally so, since the point of this submission is the experience, not a production banking system.
Challenges we ran into
The biggest challenge was scope. The full CORTEX vision — three complexity layers, a hidden Core, recursive depth inside its top features — is large enough to be a genuine product, not a two-week build. Rather than build all of it badly, we made a deliberate call early on: perfect the Surface layer completely, and treat Deep and Core as designed-but-unbuilt roadmap. That discipline is itself part of the project's story — a beautiful, fully-working slice beats a sprawling, half-working whole.
Technically, the hardest problem was making the world feel physically inhabited rather than merely decorated: getting the first-person camera, the "approach and auto-boost" logic, and the Orbit hub-fill behavior to all feel like one continuous, weighted world instead of a series of disconnected animations. A late addition — touch controls (virtual joystick + drag-to-look) — was necessary once we realized keyboard-and-mouse navigation locks out anyone without a laptop, which mattered a lot given how much of this build happened away from a full development machine.
Accomplishments that we're proud of
A fully working, end-to-end first-person experience: splash, hyperspace entrance, the Void, multiple real Orbit-based features, the Wisp, and a clean exit — all deployed to a live public link, not just a local demo. We're proud that the core thesis — speak what you want, and the world responds — is something you can actually walk into and try, right now, rather than something we're only describing.
What we learned
That immersion is a discipline, not just an aesthetic. Early in design, we deliberately drew a line between building something usable (which respects the user's time and always makes leaving easy) and something engineered to be sticky (which manipulates attention for its own sake) — and held to that line throughout, even when "make it more addictive" would have been the easier design shortcut.
What's next
- The Deep layer: a five-level staircase where task complexity — not menu depth — determines how far down a user descends, each level more elaborate and visually rare than the last
- The Core layer: a small set of the app's most powerful features, deliberately hidden — visible only as paths, not landmarks — entered through the most dramatic transition in the app: the current world breaking apart as a new one takes root
- Real account integration, replacing the current mock dataset
- A more capable intent engine, layering lightweight on-device prediction on top of the current keyword system
- Expanded touch/mobile support beyond the current baseline, so the full experience is as native on a phone as it is on a desktop
Log in or sign up for Devpost to join the conversation.