Proactive intervention
Detects when a student is stuck or stalled, and starts the conversation first.
Until now, agents have done people’s work for them. The work got done, but the people stayed the same. We go the other way. We get people to do it themselves, and we make them more capable for it.
Who we are
Same AI agents. A different thing they change.
Positioning
Software, not hardware
Neuralink and Angel Robotics augment people with hardware. We do it with software.
Acts before being asked
An agent that steps in before it is called. The category defined by Google’s YETI research, and the name of our TIPS program.
JITAI
A policy that learns when and how to intervene. A field established by Susan Murphy at Harvard in mobile health.
First implementation
Our intervention policy, first built for the SAT. 217 students study with Eden today.

Market
What parents spend on human tutors every year.
People have always been the most powerful augmentation.
But because they are people, every problem follows.
The gap
One-on-one is expensive, bound to a tutor’s schedule, and only as good as whoever teaches. Video lessons and chatbots cut the cost, but they only work when the student shows up first. Half a solution.
Product

It reaches out before the student asks.
It notices the moment they get stuck, and gets them moving again.
Eden reads intervention signals in how a student works, and opens the conversation at the right time.
If 3x + 5 = 20, what is the value of x?
Capabilities
The subject is always the student. Eden doesn’t do the student’s work. It builds what the student can do.
So students start, and don’t stop
Detects when a student is stuck or stalled, and starts the conversation first.
Remembers past work and conversations, and carries them into the next intervention.
Catches early signs of drop-off and brings the student back to learning.
In production
Since launching in SuperfastSAT in April 2026, three of the seven capabilities are running in real classes.


Technology
Just-in-time adaptive intervention (JITAI) is a field Harvard established in mobile health. We built it for learning, as an LLM agent.5
Propose. The student moves. Measure the change. Every turn of the loop sharpens the policy.
Research foundations
Product roadmap
Parents won’t accept people disappearing from lessons overnight. So we start with coaches. As the agent takes on more domains, from three to seven, the human share and the cost both shrink.
Coaches and Eden teach together. Students per coach are already rising.
Eden grows to all seven domains, reducing the human share and the cost.
Tutoring that moves students, without a human tutor.
Traction
Since July–August 2026, renewal revenue has exceeded new revenue. Students and parents are choosing us again.
Team
People who grew tutoring platforms and AI learning products, together again to build tutoring without a human tutor.


Next voyage
Getting people to move is the same problem outside learning. We started where the policy can learn the fastest.