WEBVTT

NOTE English captions transcribed from the approved V26 film and checked against the production narration.

00:00:00.000 --> 00:00:03.440
My father was living on his own when he suffered a stroke.

00:00:03.440 --> 00:00:10.930
He was found in his yard the next morning, just a few feet from a canal.

00:00:10.930 --> 00:00:16.770
We believe he may have been outside, unconscious, for as long as 10 to 12 hours.

00:00:16.770 --> 00:00:21.810
When he was discovered, he was disoriented and had no idea what was happening.

00:00:21.810 --> 00:00:28.130
What stayed with me was how long he had been alone without any of us knowing he needed help.

00:00:28.130 --> 00:00:31.090
My father has always been an independent person.

00:00:31.090 --> 00:00:35.010
I wanted to find a way to support him without taking that independence away.

00:00:35.890 --> 00:00:40.370
I wanted to build a real product to help my father and others like him.

00:00:40.370 --> 00:00:43.090
That’s why I brought together an AI product team.

00:00:44.050 --> 00:00:49.410
BMAD was the foundation. I added additional specialist agents and reusable skills.

00:00:50.290 --> 00:00:54.530
Through context engineering, product agents questioned me about the pain point,

00:00:54.530 --> 00:00:57.170
suggested options, and worked through them with me.

00:00:59.260 --> 00:01:02.140
Those decisions became the context guiding the team.

00:01:03.340 --> 00:01:06.620
Research agents compared products and identified gaps,

00:01:06.620 --> 00:01:09.420
shaping personas, user journeys, and flows.

00:01:10.220 --> 00:01:14.300
My design loop preserves consistent interactions and visual design.

00:01:14.300 --> 00:01:16.620
It also flags generic AI styling.

00:01:17.340 --> 00:01:21.500
Independent review agents flag code and potential security issues.

00:01:21.500 --> 00:01:25.260
Engineering agents make fixes and regression tests check them.

00:01:25.260 --> 00:01:27.980
Alongside the native app, I also built a web app.

00:01:28.540 --> 00:01:32.220
Agents use Playwright to check its browser flows for usability friction

00:01:33.260 --> 00:01:35.180
and automated accessibility checks.

00:01:37.250 --> 00:01:39.170
For new features and enhancements,

00:01:39.170 --> 00:01:43.570
agents debate options and refine a recommendation, then send me their conclusion.

00:01:46.240 --> 00:01:48.160
I decide whether to proceed.

00:01:48.160 --> 00:01:52.640
If I approve, a lead agent implements it in the codebase through the connected harness.

00:01:56.290 --> 00:02:01.810
The result is NearHalo, a native app live on the Apple App Store and free to download.

00:02:02.130 --> 00:02:05.490
It brings together available Apple Health data from Apple Watch,

00:02:05.490 --> 00:02:09.970
including vitals and recorded fall events, location, and geofencing.

00:02:09.970 --> 00:02:13.730
It can alert caregivers when readings cross caregiver-set thresholds

00:02:13.730 --> 00:02:15.890
or someone leaves a designated area.

00:02:15.890 --> 00:02:27.200
The product development lifecycle doesn’t stop at release.

00:02:27.200 --> 00:02:30.080
A product also needs a go-to-market strategy.

00:02:32.080 --> 00:02:35.200
I created an agent that used the product repository’s context

00:02:35.200 --> 00:02:38.240
to build the marketing website in less than one day.

00:02:38.640 --> 00:02:43.120
It captured product screenshots and wrote UX copy from our established personas.

00:02:43.120 --> 00:02:48.400
The context carried forward, from understanding the problem to bringing the product to people.

00:02:48.400 --> 00:02:54.080
I can’t change what happened to my father,

00:02:54.080 --> 00:02:58.320
but I can turn that experience into something other families can use.

00:02:59.120 --> 00:03:04.240
That’s why I built my AI product team.

00:03:05.760 --> 00:03:08.240
And that's why I built NearHalo.

