It’s easy to get lost in big AI headlines and vague promises. But on Demo Night, a room full of builders was laser-focused on one thing: real, working AI. Not “someday” promises, not endless roadmaps – actual demos solving real problems right now.
Three teams took the stage and showed us where AI is headed next: immersive game-based edtech, agentic edge security, and adaptive learning that finally makes personalization a reality. These founders didn’t just talk theory; they put it into practice. They showed what’s possible when AI gets out of the lab and into the world.
Megaminds: Turning learning into a game
Eric Tao and the Megaminds team are reshaping edtech by meeting students exactly where they are – inside games like Roblox and Minecraft. This is immersive, game-based learning at its core.
“Teachers are more overwhelmed than ever. Test scores still haven’t returned to pre-pandemic levels. And schools don’t have the tools to turn things around at scale. But that’s finally changing,” Eric said.
Megaminds’ AI tutors are designed to act like real teachers, not chatbots. “They don’t just answer questions. They engage, they guide, and they adapt in real time, just like a great teacher does.”
The results? In a Florida pilot, no ninth graders were at grade level in January. Ten weeks later, after daily sessions with Megaminds, 67% reached grade-level proficiency – more than three times the improvement of the traditional group.
“Our platform gives students a chance to level up by practicing, failing, and trying again. In education terms, this means it’s a safe space for productive struggle that drives deeper learning.”
Megaminds isn’t just another edtech tool – it’s adaptive learning designed for the Roblox generation.
CarbonDex: Giving infrastructure its own AI brain
Joseph Welsko from CarbonDex wants to move beyond passive monitoring and into true agentic intelligence for critical systems. This is edge AI and edge engineering in action.
Joseph Welsko from CarbonDex talked about bringing intelligence directly to the edge – to substations and remote sites where risks actually start. By decentralizing control, their system can predict and prevent failures before they spread across critical infrastructure.
“One of the biggest challenges of controlling something like the power grid is the time it takes for information to move and how quickly we can dispatch a resolution from our central control systems,” Joseph explained. “By moving that to substations, to the places where the risks are generated, we can predict those risks and prevent them before they have time to propagate into system-wide failure.”
CarbonDex’s system, CarbonSentinel, runs entirely on the edge – even on simple Raspberry Pi devices. Joseph highlighted, “This little Raspberry Pi right here can run not one, not two, but 10 CarbonSentinel models on different data streams simultaneously.”
It learns on-site, in real time, using spiking neural networks and liquid state machines – no cloud connection or massive GPU clusters needed. As Joseph put it, “This wasn’t rules mapping, or template engineering, or some pre-trained model. This was pure, adaptive, agentic intelligence. A system that learns from scratch, completely offline, without expensive compute.”
It’s a bold shift toward making critical systems proactive, self-learning, and ready to act independently.
Context: Understanding what people actually know
Dror Margalit and Soumya Gupta from Context are going after one of education’s biggest gaps: we don’t really know what people know until they fail.
“We’re spending $600 billion annually without actually knowing what people learn until they actually fail,” Dror said.
Context offers adaptive learning tech that maps each student’s understanding as they move through the material. Instructors can build their own courses, control the style, and then let AI personalize it on the fly.
“The AI knows what did the student actually learn? Did they actually learn what they were supposed to based on how the course creator defined it? And then the AI can improve it over time.”
Whether in K-12 classrooms or corporate training, Context pushes adaptive edtech from a marketing buzzword to an actual, measurable tool. Instead of guessing if training “worked,” educators finally get a real-time picture of each learner’s knowledge map.
Where AI gets real
From immersive, game-based learning worlds to agentic edge security to deeply personalized education, AI is stepping out of the lab and into everyday systems. Megaminds, CarbonDex, and Context each displayed their innovative products to an engaged audience of our tech community.
Want to see it live next time? Check out our upcoming AI Demo Nights and get a front-row seat to what’s actually possible.