Bounded
One experience at a time. No app store, no notifications, no background processes competing for attention. You choose what runs.
About the platform
Atlas started with a simple observation: the devices we give children are designed to capture attention, not support it. We decided to build our own.
Phones and tablets are built to keep you looking. Variable rewards, infinite scroll, autoplay, "one more." The dopamine loop is the product. Handoff becomes a battle because the device was never meant to be handed off.
That's not a parenting failure. It's an engineering decision made by companies whose revenue depends on screen time. We wanted hardware where the engineering decisions serve the learner instead of the ad model.
So we built Atlas: compute platforms with open firmware, no app store, no tracking, and no incentive to keep you hooked. The device does what you configure it to do and nothing else.
Atlas is a family of compute devices — handhelds, tablets, sensor modules, and a local appliance — that run the Castalia learning stack. The firmware is open. The hardware is purpose-built. Every form factor shares the same software stack and talks to the same ecosystem: iNQ Cards, terrAIn robotics, Atlas content libraries.
For young children, Atlas runs curated activities in kiosk mode — one activity at a time, large touch targets, no ads, no text input. At rest time, Nap Mode turns the screen into a warm amber nightlight. If the device detects sustained wakefulness, it alerts you instead of escalating on its own.
For older learners and field use, the same hardware runs iNQ Card scanning, terrAIn vehicle control, content creation tools, and connects to the Inquiry Appliance for local AI inference. The platform grows with the user — different firmware profiles, same device.
Every decision in the Atlas stack traces back to these four constraints.
One experience at a time. No app store, no notifications, no background processes competing for attention. You choose what runs.
On-device processing only. Sensor data discarded within 100ms. No telemetry, no analytics SDK, no server to phone home to. The Inquiry Appliance keeps everything on your network.
Firmware source available. Flash your own builds. Inspect what's running. We don't lock the bootloader and we don't void the warranty for tinkering.
Drop-tested housings, recessed glass, matte shells. These are instruments designed for years of daily use — by toddlers, by students in the field, by teachers who need something that survives a classroom.
For the youngest users — a rest mode that actually respects sleep.
Nap Mode uses movement and breathing cues — no facial recognition, no emotion labeling. Detection confidence is always treated as uncertain: when in doubt, the device holds the current state. The screen becomes a warm nightlight; sound is capped; voice prompts, if enabled, are whisper-soft: "You're safe." "Night is here." "I'm holding the quiet."
Stable breathing, minimal motion — the device holds the quiet.
Settling after bedtime — gentle guidance.
Micro-waking detected — repair without startling.
Sustained wake signals — parent handoff when needed.
Atlas hardware runs the Castalia learning stack. The same iNQ Cards work on every device (DS and SP require the Sense board for card scanning). terrAIn robots connect over the same firmware. The Inquiry Appliance serves content and AI to the whole fleet. It's one system with multiple form factors — not isolated products.
castalia.institute — the broader platform. terrAIn — the robotics arm.