There is a particular kind of moment when most software asks too much.
A craving is loud. Anxiety is accelerating. Shame is narrowing the field of view. A person feels alone, overwhelmed, or uncertain whether the next decision will make the situation harder.
That is not the moment for an onboarding sequence, a diagnostic questionnaire, a motivational dashboard, or a demand to explain everything in complete sentences.
It may be the moment for one smaller question:
What would make the next few minutes a little safer?
That question became MARGIN, a private, no-account web prototype for people navigating substance-use pressure and mental-health distress at the same time.
MARGIN does not promise treatment. It does not predict relapse, monitor a crisis, or grade recovery. It helps a person reduce one difficult moment into one manageable next decision.
The product category was part of the problem
Addiction and mental health are often separated by institutions, funding streams, service directories, and product categories. Lived experience is rarely so tidy.
A craving can follow panic, conflict, sleeplessness, grief, isolation, or a depressive period. Substance use can intensify anxiety or low mood. A person may not know which label belongs at the top of the screen, and they should not need to choose the correct institutional doorway before receiving basic support.
SAMHSA’s guidance on co-occurring disorders emphasizes integrated care for mental-health and substance-use conditions. I wanted the product architecture to reflect that reality from the first interaction.
The opening question is not “What is your diagnosis?” or “How many days have you been sober?”
It is: What feels loudest right now?
The choices are ordinary and non-diagnostic:
- A craving
- Too much at once
- Feeling low
- Feeling anxious
- Feeling alone
- Not sure
“Not sure” matters. Uncertainty is not a form error. It is a valid state from which to begin.
Recovery should not become a red reset screen
Many products use streaks because streaks are easy to understand, easy to display, and effective at bringing people back to an app.
But a streak carries a theory of change. It says continuity is the central measure and interruption is a reset. In recovery, that visual grammar can turn a difficult event into the apparent destruction of everything that came before it.
I did not want MARGIN to make that claim.
The prototype has no sobriety clock, recovery score, leaderboard, lost-progress animation, or “clean versus dirty” language. The CDC recommends person-first, non-stigmatizing language, and the product extends that principle beyond copy into the structure of the interface.
MARGIN recognizes three kinds of progress:
- Creating a pause before a decision
- Changing something in the immediate environment
- Bringing another person closer
Those are not clinical outcomes. They are actions a person can choose when a larger promise may feel impossible.
Making a margin
After a user names the pressure, MARGIN asks what “safer” should mean for the next ten minutes.
The product then creates a three-part plan:
- Body — a small physical or sensory action
- Space — one environmental change that may lower pressure
- Connection — one way to involve another person
The user can replace any action. Nothing is framed as a command, and no breathing exercise is mandatory. For some people, focusing on breathing or closing their eyes can feel uncomfortable or unsafe. A trauma-informed interface has to preserve more than good intentions; it has to preserve choice.
The plan stays on one screen beside an optional ten-minute timer. The timer can pause or reset. It does not announce every passing second or turn the moment into a performance.
At the end, the reflection asks: a little easier, about the same, or harder?
“About the same” is not recorded as failure. “Harder” offers a direct path toward more support. The prototype does not build a private behavioral profile or claim that a single action caused the change.
The hardest button was the human one
“Contact someone” sounds simple until someone has to decide what to say.
During a high-pressure moment, reaching out can require admitting need, predicting another person’s reaction, and translating a complicated experience into a message. That is a lot of work before support has even begun.
MARGIN’s quiet relay offers an editable sentence:
I’m having a high-pressure moment. Can you stay with me for ten minutes? You don’t need to fix anything.
The wording is deliberately bounded. It makes a specific request, gives the other person a role, and removes the expectation that they solve the entire situation.
MARGIN never sends the message automatically. It does not select a contact, inspect a contact list, monitor a reply, or pretend a human is present when no one is connected.
Crisis design cannot be decorative
A persistent Urgent Help control is visible throughout the experience. It does not pulse, flash, or wait behind an account wall.
The dialog separates four situations:
- A possible overdose or immediate medical emergency
- Emotional distress or a substance-use crisis
- Treatment and referral support
- Withdrawal that may require medical attention
That separation is important. A possible overdose calls for a different response than a request to find treatment. The prototype summarizes immediate overdose steps from CDC overdose-response guidance, including calling 911, administering naloxone if available, helping with breathing, placing the person on their side, and staying with them.
For emotional distress or a substance-use crisis in the United States, MARGIN links to the 988 Suicide & Crisis Lifeline by call, text, and chat. The interface says that a person does not need to be suicidal to contact 988.
Treatment routes link to the SAMHSA National Helpline and FindTreatment.gov.
The withdrawal note is equally direct: the prototype should not be used as a plan for suddenly stopping alcohol, sedatives, or prescribed medication. Severe withdrawal can require emergency medical care.
Every urgent route also states the boundary: MARGIN does not monitor the page, dispatch help, diagnose a condition, recommend medication, or provide a self-detox plan.
Privacy by subtraction
Health-related privacy is often presented as a policy document after the product has already decided to collect everything.
MARGIN starts from subtraction.
There is no account. The current prototype keeps selections only in page memory, so refreshing or closing the page clears them. It does not use local storage, session storage, advertising identifiers, analytics, pixels, or session replay.
That does not make the experience magically anonymous. Hosting providers may retain ordinary request or security logs. The privacy dialog says so.
The quick-exit button opens a neutral weather site and replaces the current history entry where the browser permits it. It also explains what that action cannot do: erase other browser history, device traces, downloads, network logs, or evidence visible to someone with access to the device.
Privacy is not a slogan here. It is a list of things the product deliberately does not ask for.
I used agents, but I did not automate judgment
I developed MARGIN with a team of specialist AI agents, each assigned a different discipline:
- Addiction and mental-health product strategy
- Clinical-safety research and crisis boundaries
- Trauma-informed UX and accessibility
- Visual identity and social presentation
The agents proposed competing concepts, identified risky language, examined how urgent-support routes should differ, and tested the product thesis against existing recovery-app patterns. I made the final product decisions, reconciled conflicts, wrote and revised the experience, implemented it, and tested the working prototype.
I bring what the agents cannot: lived experience as a person in recovery from addiction and alcoholism, together with a trauma-informed perspective. That experience shaped the product’s refusal to use shame, streak loss, forced disclosure, or surveillance as motivation. It is one person’s perspective, not a substitute for broader paid co-design or professional review.
This collaboration made the work faster and more multidimensional. It did not turn the agents into clinicians, crisis counselors, or people with lived experience.
That distinction is essential. SAMHSA’s peer-support competencies emphasize voluntary, person-centered, relationship-focused, and trauma-informed support. A production version of MARGIN should not merely cite those values. It should give paid governance authority to people with lived and living experience and undergo formal clinical, crisis, privacy, legal, and accessibility review.
What this prototype can honestly claim
MARGIN is a working product prototype. It is responsive, keyboard accessible, reduced-motion aware, and usable without an account. Its crisis and treatment links are visible, its privacy boundaries are explicit, and its central loop can be completed in seconds.
It is not clinically evaluated. It is not a medical device, treatment program, crisis service, or substitute for professional or emergency care.
The prototype demonstrates a design argument rather than a health outcome:
A digital support tool can help make room for a choice without demanding an identity, a streak, a diagnosis, a history, or a reason to keep scrolling.
If MARGIN eventually becomes a public service, I would measure whether people understand what is stored, reach a chosen action quickly, connect with human support when they want it, and experience the interface as increasing rather than reducing agency.
I would not optimize it for daily active use.
The best version of MARGIN may be the product someone leaves after two minutes because they found the next step they needed.
Open the MARGIN prototype or read the complete project case study.
Disclosure: I bring my own lived experience in recovery and a trauma-informed perspective to MARGIN. I also used specialist AI agents for research, product strategy, safety analysis, UX critique, visual development, and implementation support. The prototype remains an independent concept and has not received formal clinical review or broader paid lived-experience governance beyond my perspective.