AHA← Back to site
Open the live demo →
1 / 17
Story walkthrough · for video tutorial01 / 17
AHA
9:41 AM · Intro to Statistics · 142 present

Everyone looks
equally present.

A large online lecture. Every camera a black tile. The instructor is talking to a wall he can’t read, and can’t tell who’s actually still with him.

Open the live demo
Agung · Wesley · Kashvi · Yangchen · Sheila · Karlee
Group 1 · Technology Product Management · Columbia University · Summer 2026 · aha-moment-nine.vercel.app
JK
PN
ML
AP
MT
JR
AS
RC
142 cameras. No way to tell who’s learning. (hook video slot)
Problem statement02 / 17

Attendance is visible. Learning isn’t.

In live online lectures of 100+ students, instructors have no reliable way to know when students feel unseen, confused, passive, or disengaged. It stays invisible until it shows up as a poor exam or a dropped course, too late to act.

6
in-depth user interviews across learner types, synthesized into four personas and one root problem.
What research changed
Users rejected our first idea (a learner-side completion tool). The problem itself moved: from fixing content to detecting the moment of disengagement.
Root insight

The problem was never content quality. It’s that nobody notices the moment a student checks out.

User persona · before AHA03 / 17

Meet the instructor living that gap.

MT
Michael Torres
Adjunct lecturer · 100+ student intro course
“In a physical classroom, I can read the room without thinking. Online, everyone looks equally present, whether they're with me or not.”
  • , He needs to know, in real time, who's confused or checked out.
  • , By the time a quiz shows him who was lost, the session is already over.
  • , Every time he tries to re-engage the room, it's a guess, he can't tell if it worked.
  • , Cold-calling is his only live signal, but it breaks his teaching flow.
Vision statement04 / 17
AHA

So what would it take to close that gap?

Restore the ability to read a room the moment class goes online. Turn attendance from a task-driven obligation into a momentum-driven learning experience, by giving instructors the awareness a physical classroom gives them for free. The adaptive layer comes later, and only after the diagnostic has earned it.

Value proposition05 / 17

A co-pilot for the classroom, answering his pain directly.

1Live Engagement Dashboard

A real-time picture of the class in three states, so the instructor knows where to direct attention right now.

✓Answers: “He needs to know, in real time, who's confused.”
2AI Instructor Copilot

Suggests the next best action when a group drifts: a poll, a break, a check-in. The instructor stays the pilot.

✓Answers: “Every re-engagement is a guess, and cold-calling breaks his flow.”
3Fits Existing Tools

Zoom API and Brightspace integration. No new platform to learn; AHA lives inside the class he already runs.

✓Answers: “No appetite for another tool to manage.”
4Trust by Design

On-device processing, disclosure at session start, session-level deletion, and the Two-Witness Rule.

✓Answers: “It can't feel like surveillance to students.”
Solution statement06 / 17

AHA is an attention intelligence layer on the live class.

Instructor dashboard

Real-time engagement per student, an aggregate class score, and a one-click private ping.

Instructor copilot

When the class dips, it suggests a recovery move: a poll, a cold-call, a recap.

Facial + behavioral signal

Detects eyes-away, drowsiness, and confusion. All processed locally, never transmitted.

01 Signals
Facial + behavioral, on-device
›
02 Two-Witness
Only when two signals agree
›
03 State
Engaged / Drifting / Disengaged
›
04 Dashboard + Copilot
Instructor sees, Copilot suggests
›
05 Instructor acts
Ping, poll, or recap
MVP · the product, live07 / 17

What the instructor sees.

The working prototype at aha-moment-nine.vercel.app, the same loop, running in the browser right now.

aha-moment-nine.vercel.app/instructorLIVE
Intro to Cognitive Science · Section 002
JK
Jordan Kim
● Engaged
PN
Priya Nair
● Drifting
ML
Marcus Lee
● Engaged
AP
Aisha Patel
● Disengaged
JR
Jonah Rim
● Camera Off
AS
Ali Song
● Engaged
Class engagement71%
COPILOT
Engagement dipping, suggested: a quick poll or two-minute check-in.
Private ping → Aisha
“Still with us? Let’s refocus.”
Real-time state per studentLive class-engagement scoreCopilot + one-click private pingOpen the live demo →
Trust & privacy08 / 17

Isn’t this just surveillance? We only see engagement, not students.

For instructors
A co-pilot, not an autopilot

Every intervention is one click and instructor-approved. The Two-Witness Rule means a facial signal alone never fires an alert.

For students
Privacy by design

On-device processing. Zero raw video stored or transmitted. Only the engagement state ever reaches the instructor.

For admins & IT / legal
Built for trust & compliance

Per-student baselines tested for disparities across disability, neurodivergence, lighting, and skin tone; consent tracking, IRB & FERPA readiness.

And we give students a voice, Live Pulse
Only “I need help” ties to a student and auto-suggests a Private Ping. The other three stay class-level aggregates (e.g. “23% confused”), so most of Pulse never identifies anyone. It validates; it never triggers an alert on its own.
I understandI'm confusedPlease slow down→ I need help
User persona · after AHA09 / 17

Torres, again. What does his class look like now?

  • , Now he knows, in real time, who's confused or checked out.
  • , The session isn't over before he finds out who was lost.
  • , He's no longer guessing, he knows if the intervention worked.
  • , Cold-calling isn't his only signal anymore, and his flow stays intact.
MT
Michael Torres
Same class. Same 142 students.
“Online teaching still isn't the same as being in a physical classroom. But now, I can read whether they're with me or not.”
Key success metrics10 / 17

One North Star, three KPIs, hard guardrails.

North Star Metric
High Engagement Minutes per Student per Session

The delta between AHA-on and AHA-off sessions, normalized per student and per session so a 200-student lecture and a 40-student seminar compare fairly.

Guardrails & gates
▲False positives below 15% overall and below 25% for every student subgroup (fairness, per Buolamwini & Gebru).
▲Student consent rate above 70% before scaling.
▲Learning validity: engagement signals must correlate with real quiz outcomes before curriculum analytics ships.
Flag Precision Rate

Share of alerts instructors judge correct. Validated in Pilot against Live Student Pulse as ground truth.

Instructor Action Rate

Share of alerts that lead to a real action. An unused alert is a failed alert.

Weekly Active Instructor Rate

Repeat weekly use. Engagement minutes mean nothing if instructors stop showing up.

See the metrics dashboard →
Go-to-market strategy11 / 17

Instructors create demand. Institutions sign contracts.

1
Bottom-up: Instructors

Free semester-long design-partner pilots at 3 universities via warm academic channels. Advocacy before any sale.

2
Platform: Zoom

The Zoom API integration puts AHA inside the tool instructors already open every day. Friction drops; visibility compounds.

3
Top-down: Institutions

EDUCAUSE-type IT & procurement conferences reach the legal, privacy, and IT approvers instructor enthusiasm can't.

Pricing model: instructor tools stay free to drive adoption; the paid tier is the institutional dashboard, billed per seat at the department level.
~$1M
first-year ARR goal
Tiered pricing, budget, and ARR benchmarks → appendix.
The AHA moment12 / 17

From black tiles to the AHA moment.

We started at a wall of black tiles, an instructor who couldn't read his own room.
Now he sees it in real time, acts while it still matters, and knows if it worked.
Two witnesses before one alert. Consent before scale. Validity before analytics.
Give instructors back the room.
Try it live →aha-moment-nine.vercel.app · Thank you · Questions welcome
AppendixCompetitive landscape13 / 17

Where AHA wins, and what we refuse to copy.

Player
Focus
Segment
Signal approach
What we learned
GoGuardian
Classroom management
K-12
Device monitoring
Win by owning ONE narrow segment
Proctorio
Exam proctoring
Higher ed
Coverage-first surveillance
Coverage without precision creates harm
Engageli
Engagement platform
Higher ed
Requires its own platform
Replacing Zoom is too much friction
Kortext / Adaptemy
Content analytics
Higher ed
After-class data
Insight after class is insight too late
AHA
Live disengagement detection
100+ student live courses
Two-Witness, on-device
Precision + consent as the wedge

Key differentiators: the Two-Witness Rule (a delighter no single-signal competitor offers), privacy-first on-device processing, a cameras-off behavioral fallback, and alerts designed for instructor action rather than student punishment.

AppendixAdapting under pressure14 / 17

Three twists that shaped the product.

1Module 4
Users rejected the original idea
Impact

Our learner-side completion tool failed Cagan's value risk: people didn't want it badly enough.

Decision

Pivoted to the instructor side: detect the moment of disengagement instead of fixing content.

Result

Redefined the problem, the persona, and the entire MVP.

2Module 6
Mid-project team reshuffle
Impact

Team changed after the topic and early interviews were set; new members lacked discovery context.

Decision

Split ownership by strength and documented decisions so context transferred fast.

Result

The competitive analysis and buyer-side thinking exist because of the reshuffle.

3Module 10
Stakeholder's late feature request
Impact

A senior stakeholder proposed Live Student Pulse: students self-report their own state.

Decision

Accepted, but scoped as a third independent validation signal, not a trigger.

Result

Pulse becomes Pilot-phase ground truth for Flag Precision Rate, without letting one click override two agreeing signals.

If Two-Witness is how AHA watches, Pulse is how it listens.
AppendixPricing & unit economics15 / 17

The commercial detail behind the one-liner.

Tiered pricing
$200–300
per instructor / year, billed per seat at the department level. Instructor tools stay free.
6-month GTM budget
$190–280K
Sales/BD $80–120K · Pilot onboarding $60–90K · Legal/compliance $30–40K · Conferences $20–30K
ARR benchmarking
2–3%
marketing as a share of ARR, well below the 15–25% seed-B2B benchmark. Total GTM spend: 19–28% of first-year ARR.
See the full GTM plan →
AppendixQ&A · when signals disagree16 / 17

We never recalibrate off one mismatch.

Live Pulse is a third, independent signal, it validates the Two-Witness engine but never triggers an alert on its own. So we apply Two-Witness logic to the fairness check itself: one signal can never override another alone.

Self-report “I understand” + outcomes stay strong
The AI read is the outlier

Self-report and quiz results agree the student is fine, so a Drifting flag is likely the miscalibration. We treat their baseline as a recalibration candidate, but only if the pattern repeats across sessions, never off one class (which could just be prior knowledge).

Self-report “I understand” but quiz scores drop
The self-report is unreliable

Not the AI being wrong, not the student lying, engagement ≠ learning, the exact gap we named. The click can't be trusted as a learning signal, so it never recalibrates anything.

The rule

Recalibrate a baseline only when the same mismatch repeats across sessions and a third signal (quiz / comprehension) agrees. One signal never overrides two.

The framing

Human-in-the-loop, where the human is the student too.

Bonus

Closes the accessibility gap, and gives a faster second signal alongside quiz data to check detection.

AppendixReferences17 / 17

Sources & frameworks.

01Buolamwini, J., & Gebru, T. (2018). Gender shades: Intersectional accuracy disparities in commercial gender classification. PMLR, 81, 1–15.
02Cagan, M. (2018). Inspired: How to create tech products customers love (2nd ed.). Wiley.
03Kano, N., Seraku, N., Takahashi, F., & Tsuji, S. (1984). Attractive quality and must-be quality. J. Japanese Society for Quality Control, 14(2), 147–156.
04Course readings: Pilaniwala (2024); Rutkowski (2022); Sreenivasan & Suresh (2024).
05Primary research & team deliverables: 6 user interviews, Competitive Landscape (incl. TAM/SAM/SOM), IBMC, User Stories, 6-Month Roadmap (RICE, Kano, gates), Success Metrics Dashboard, GTM Roadmap, working prototype at aha-moment-nine.vercel.app.
06Zoom (2026), Nonverbal feedback features; Mentimeter (2026), Live polling.
AI disclosure: portions of research synthesis, drafting, and prototype development were assisted by AI tools; all analysis and decisions are the team’s own.