STOA
AI-guided anxiety intervention, exactly when you need it.

Bottom line up front (BLUF) · 1 minBottom line up front
- Apps built to help during an acute anxiety attack fail because they wait for the user to open them.
- In the middle of a panic-inducing situation, the ability to reach for help and calm yourself quickly slips away, so most people never pick up their phone at all.
- Avoidance brings short-term relief, but it keeps the cycle going.
- I designed STOA, a phone and smartwatch app that makes the first move during an acute anxiety attack.
- Its smart band, or the user’s own smartwatch, is designed to sense rising stress levels.
- The app then starts a short exercise based on cognitive behavioral therapy (CBT) right away.
- My role
- Solo designer
- Timeline
- Winter 2026, six weeks from research to prototype
- Team
- Just me, with Claude as an AI design partner
- Took it from research to a tested prototype
- Research: a synthesis of 42+ sources
- A 109-token design system, and every phone and watch screen
- A Wizard-of-Oz test of the band, with its alert triggered by hand
- 01Built the exercises on CBT, and used Stoicism only as their voiceFrom a live design critiqueThe early concept leaned on Stoicism more than the clinical evidence supports.
So I built the prompts on CBT and kept Stoic quotes for their tone.
- 02Replaced a yes-or-no check-in with five honest optionsFrom an AI critique I ran on the screensEnding the exercise with “I feel calmer” assumed it worked, and left no path for someone who felt worse.
So I added a post-attack check-in with five answers, including “still anxious.”
- 03Left out heart-rate charts, and didn’t prototype the AI’s decision-makingFrom my researchReliability matters more to people than a long list of features.
So STOA stays narrow and, once it knows someone, checks in only when needed.
In the midst of a panic attack, nobody reaches for an app. STOA reaches out first.
- Intervene before the person loses the ability to think clearly
- Complete an anchoring technique, like a breathing or reframing exercise, in less than 2 minutes, in a public setting
- Zero false-alarm fatigue: detection precise enough that every alert means something
From the first spike in heart rate to a recommendation the user can trust
Try the prototype yourself.
Introducing STOA.
A three-minute walkthrough in my own words: the smart band detecting an acute anxiety attack, choosing an anchor technique, and the Stoic voice that carries the exercise.
Read the transcript
Hi, my name is Jarel. I’m a combat veteran, and I’ve had my fair share of dealing with PTSD and anxiety. I can tell you from experience that when trying to manage anxiety and panic attacks, the moment you most need help is exactly the moment you’re least likely to open an app. Every meditation product in the market waits for you to notice that you’re anxious, decide to act, and open up your phone and find the app. My product, STOA, actually flips that. It’s an AI guided intervention system that pairs a wearable with your phone, so it notices before you do. I named it for the Stoics, who taught that anxiety lives in our perceptions, not in the world. And you’ll see that philosophy show up in the product itself. Stoicism is also the ancient predecessor of cognitive behavioral therapy, or CBT, which is regarded as a legitimate goal-oriented psychotherapy that is a first-line treatment for things such as anxiety, depression, stress, and anger management.
So let’s go ahead and dive in. So go ahead and click Get Started, and this first screen is onboarding, and the onboarding is essentially just one question: What actually calms you? Is it your breath? Is it your thoughts? Is it your body? That single choice personalizes everything downstream. So today we’ll go ahead and choose that we want to reframe our thoughts, and as you can see, the welcome screen here, you have a healthy heart rate, 72 beats per minute. But look what happens! All of a sudden, you have elevated stress. The heart rate spikes to 140. STOA reaches out to you with plain language and no alarm bells, and a choice to anchor yourself in the moment. So we’ll go ahead and choose to anchor ourselves. From here, you choose how you want to anchor: calm my body, reframe my thoughts, or you can let STOA make a recommendation. This recommendation isn’t random. It’s actually drawing on your history, your chosen anchor, and what’s worked before.
So let’s go ahead and let STOA decide. All right, so it’s decided to go ahead with cognitive reframing for two minutes. So we’ll go ahead and begin. And today’s reframing is a quote from Marcus Aurelius, who was an ancient Roman emperor and perhaps the most famous Stoic philosopher. There’s a two-minute timer to meditate on the cognitive reframing. When you feel a bit more relaxed and your heart rate is hopefully slowed down, you can let the app know that you feel a lot calmer. So afterwards, STOA closes the loop by asking how you feel, and that answer trains the next recommendation. So today, we’re going to go ahead and choose that we’re feeling much better, and over time, you see the proof. 89% of sessions help you feel calmer. Overall, your panic episodes are down, and over time, you’re recovering faster.
So overall, this project taught me that the best intervention is the one that arrives before you ask. I built this tool as something that I personally needed, but that maybe one day it will help others as well. Thanks for your time and thanks for watching.
The evidence says the system has to make the first move
I synthesized more than 42 academic sources across six research questions using Elicit, an AI research tool, then read first-person accounts from people who have had panic attacks. Four findings came back, and none of them left room for an app that waits to be opened.
“Even in the midst of the darkest panic attack, part of you is still rational. The rational part is paralyzed, a helpless spectator to your madness.”
- 01
Anxiety sensitivity is the real problem.
People fear the sensations of anxiety, the racing heart, the tight chest, more than whatever causes them.
- 02
Help timed to the body beats help on a schedule.
Exercises triggered by a person’s own stress thresholds work significantly better than scheduled reminders or random notifications.
- 03
Consumer wearables can detect anxiety accurately.
Smartwatches already do it well, and detection models tuned to one person beat generic ones after one to two weeks of learning.
- 04
CBT is the framework. Stoicism is the voice.
Stoicism is the ancient philosophy of separating what you can control from what you can’t. Modern CBT draws directly on it, as its founders Albert Ellis and Aaron Beck acknowledged, so quoting the Stoics isn’t decoration. It points back to where the app’s techniques came from. The evidence supports reframing thoughts, changing behavior, and calming the body, so STOA uses all three.
Detection can’t be tested on a screen, so I tested it on a wrist
I built a band with an LED and controlled it by hand while someone wore it. It lit amber when I decided their stress had spiked, and green once the exercise brought them back. It was a Wizard-of-Oz test, with me as the person behind the curtain. This rig stands in for STOA’s smart band, made for people who don’t own a smartwatch. It pairs with their phone, which shows the alert and runs the exercise. In testing, the exercise finished inside its two-minute limit and the experience felt the way it should, so the flow stayed as designed. Next, I’d test a band that fires when nothing is wrong, to check the zero false-alarm goal. Once STOA is live, three numbers would show whether it helps: how often people dismiss an alert, how quickly they start the exercise, and how often they answer “Still anxious.”
The rig: a strap, an LED, and a microcontroller I drove by hand. Amber stands in for detection. The wearer feels the alert before they name the feeling. Green is the return to baseline, the thing the whole product is trying to produce.
Every part of the design system had to work for someone mid-attack, in public, with shaking hands
So I stripped the interface to almost nothing and checked every color pair against accessibility contrast standards.
- 109Design Tokens37 primitive, 24 semantic, 48 component
- 8Type StylesCormorant for reflection, Jakarta Sans for action
- 4Brand ColorsNo red, nothing that could read as alarm
Ten colors, and not one of them is red
Red means danger, and a product that tells an anxious person something is wrong has already made things worse. The palette is built the other way: Aegean Deep to settle, Bronze for the single action worth taking. Its four brand colors, Aegean Deep, Bronze, Obsidian, and Marble White, trace to the ancient Mediterranean, where Stoicism started. The contrast checks below test each pair against WCAG AA.
- Aegean Deep#184C5Acalm · surfaces · outlines
- Bronze#A67C52the one action on the screen
- Obsidian#0F2E38headings on light
- Marble White#F7FAFBthe ground everything sits on
- Calm Blue#5A8E8Esecondary accents
- Bronze Light#D4A574rings · detection state
- Aegean Tint#E8F1F3cards on the gradient
- Pewter#5A7A82body · secondary text
- Light Stone#C7D4D8borders · unselected
- White#FFFFFFcards · sheets
- White on Aegean Deep9.46:1 · passes WCAG AA
- White on Obsidian14.30:1 · passes WCAG AA
- Obsidian on Marble White13.64:1 · passes WCAG AA
- Pewter on Marble White4.41:1 · below WCAG AA
- White on Bronze3.73:1 · passes WCAG AA for large text only
One typeface runs the interface, and another appears only for the Stoic voice
Plus Jakarta Sans handles every control and label. Cormorant Garamond marks each Stoic passage, so the change of voice is visible before a single word is read.
- Calm in any stormDisplay · 32px · one per screen
- Calm in any stormTitle · 22px · section heads
- Calm in any stormAction · 17px · buttons
- Calm in any stormBody · 16px · instructions
- Calm in any stormBody Small · 15px · supporting copy
- Calm in any stormCaption · 14px · metadata
- Calm in any stormMicro · 11px · tab labels only
- Calm in any stormReflection · 28px · the Stoic voice
Sized for hands that are shaking
Everything a user touches mid-attack is bigger than accessibility guidelines require, so a shaking hand still lands on the right button.
- 44 × 44WCAG 2.5.5 (AAA) target
- 48 × 48the STOA floor
- 72 × 72tab bar item
Smartwatch owners get STOA on the wrist, so the first alert never waits on a phone
After the exercise, STOA checks in, then keeps a record you can look back on
What designing for a crisis taught me about AI
- 01
An AI that can’t show its reasoning can’t be trusted, least of all mid-panic.
On the recommendation screen, the question was never which exercise the AI should pick. It was whether an anxious person would believe it. Advice without a reason is easy to dismiss, and someone mid-panic will dismiss it.
- 02
What I left out shaped STOA as much as what I built.
My research found that people value a tool that works every time over one that does everything, so I kept STOA narrow. I didn’t prototype the AI’s decision-making, because that would have meant solving a problem I hadn’t reached yet. And once STOA has learned someone’s patterns over the first one to two weeks, it monitors less closely and checks in only when needed, so it doesn’t interrupt when nothing is wrong.
- 03
AI is good at noticing. It’s bad at caring.
The watch detects the spike, and STOA’s AI proposes the exercise. What actually calms someone came from research and empathy: a Stoic voice where a clinical alert would usually sit, and the right tone at 140 beats per minute.
What I would do differently
- 01
Resolve the “Still anxious” path first.
I would prototype the failure state before the success state. What STOA does after someone answers “still anxious” is the one path I left unbuilt, and in a mental health product it carries the highest emotional stakes.
- 02
Add consent before biometrics.
Onboarding jumps from choosing an anchor straight to a live heart-rate dashboard, without explaining how that data is collected, stored, or used. A product that reads your body has to ask permission before it starts listening. My research also found that privacy goes largely unaddressed across mental health wearables, so asking first would set STOA apart.
- 03
Fix the two colors that fall short of accessibility standards.
The Bronze button label (3.73:1 at 17px) and Pewter body text (4.41:1) both miss WCAG AA. Setting the label at 19px bold and darkening Pewter to #57757D fixes both without moving anything else in the system.