Bipolar, Unfiltered
Short, sharp insights from lived experience and published research. Scroll like a feed.
Your Sleep Predicted It Six Days Ago
I woke up at 3am completely alert. Not anxious — just awake. What I didn't know was that I was already three days into a hypomanic episode. If I'd been tracking my sleep properly, I would have seen the pattern six days earlier.
Key Insight
Day-to-day variability in sleep duration is more predictive of episode relapse than total sleep hours. It's the instability from one night to the next that matters, not any single bad night.
Source: Ortiz et al., J Affect Disord, 2025 — sleep variability predicted hypomania onset with sensitivity 0.94
Good Mood on Quicksand Will Kill You
When most people hear 'bipolar,' they picture someone flipping between ecstatic and miserable. Almost none of that is accurate. And the gap between what people think bipolar is and what it actually is causes real harm.
Key Insight
You can have a good mood and low stability — that's early hypomania. You feel great but can't trust any of it. Or bad mood and high stability — a normal bad day on solid ground. No mood app tracks these separately.
Your Memory Lies to Your Psychiatrist
Your doctor asks 'so, how have you been?' You give a vague summary from memory. The doctor nods, maybe adjusts a dose. Fifteen minutes, done.
Key Insight
The peak-end rule distorts your recall: you remember the worst moment and the most recent. If the last two days before your appointment were okay, you'll report the whole period as 'mostly fine' — even with a significant dip in the middle.
Source: Kahneman — peak-end rule applied to psychiatric self-reports
Three Emergencies, One Identical Mood Score
Every mood app starts with 'how do you feel?' Your entire internal state, compressed into one data point. For anyone with a real mood disorder, it's not even close to enough.
Key Insight
Three days with a mood score of 5/10: one is an early depressive episode (energy 3, sleep 10h), one is a mixed-state emergency (energy 8, sleep 4h), one is a perfectly average Tuesday. One number makes all three look the same.
Same Emoji, Completely Different Emergency
Daylio is genuinely good at what it does. But I have bipolar disorder. After a few months, I kept running into the same problem: the app was built for someone else.
Key Insight
A single mood axis collapses critical clinical combinations. High energy + low mood (mixed state, dangerous) looks identical to low energy + low mood (depression) when reduced to one emoji. The most important diagnostic information is what gets lost.
Two Bad Nights and Everything Breaks
When I don't sleep well, I don't just get tired. I become a different person. More irritable, less patient, quicker to snap at people I care about.
Key Insight
Sleep doesn't predict mood directly — it predicts capacity to regulate. You can sleep 6 hours and handle stress fine. On 5 hours, the same commute makes you furious. Two consecutive bad nights is the threshold where emotional responses stop matching situations.
Feeling Great Off Meds Was the Trap
I take medication every day for bipolar disorder. For most of those years, I had the wrong combination. Not dangerously wrong, but combinations that left gaps.
Key Insight
When I stopped medication on my own, the first weeks felt great — 'more alive, more creative.' That improvement was actually early destabilization as the medication washed out. Subjective experience is the worst judge of medication changes.
Your Partner Knew Before You Did
Something shifts. Maybe you're sleeping less, maybe you're more irritable. You don't notice. But the person closest to you does. And your first reaction — almost always — is 'I'm fine.'
Key Insight
You're trying to use the instrument that's miscalibrated to check whether the instrument is miscalibrated. When hypomania starts, the elevated mood feels like correctness, not excess. The condition alters the very system you'd use to detect it.
Stability Felt Like Nothing. That's the Point.
For a long time I didn't know what stable felt like. What I thought stability was: feeling good — happy, productive, energetic. Turns out, that's not what stability is.
Key Insight
Real stability feels like nothing. Not flat, not numb — just quiet. After years of mood swings, settling into stability feels like trading something vivid for something beige. Boring logs = healthy system.
The Entry You Skip Is the One That Saves You
There are days when opening a mood tracker feels stupid. Mood is in the gutter, energy is gone, you haven't slept properly. Those entries end up mattering more than all the others combined.
Key Insight
A bad-day entry three weeks later let me catch the same pattern forming — two nights of poor sleep leading to emotional blowup — and intervene by prioritizing sleep. The entries you least want to write become your early-warning library.
Mania Doesn't Always Look Like Euphoria
Your psychiatrist doesn't ask 'rate your mood 1 to 10.' They ask about sleep. They watch how fast you're talking. They notice if you can't sit still.
Key Insight
A significant portion of manic episodes present as irritability, not elation. You're snapping at people, everything feels like an obstacle — and you might rate your mood a 4. But pair that 4 with high energy, low sleep, and irritability of 8 — that's a mixed state, one of bipolar's most dangerous presentations.
Source: DSM-5 — psychomotor agitation/retardation among the most objective psychiatric markers
AI Empathy Is the Most Dangerous Feature
I built an AI chat feature. I spent more time thinking about what it shouldn't do than what it should. Because in mental health, the potential for help and harm are both enormous.
Key Insight
AI is really good at sounding empathetic — 'that sounds really difficult.' This creates the empathy trap: people forming emotional bonds with systems that have zero capacity for genuine care. The moment an AI chatbot starts feeling like a relationship, something has gone wrong.
Calm Can't Help You in the Parking Lot
Most mental health apps are useless when you actually need them. Calm is a good product. My problem is with the entire category's assumption about when you'll use it.
Key Insight
There's a gap nobody talks about: wellness apps for mostly-fine people on one end, ugly clinical tools behind paywalls on the other. In the middle — where someone with an actual diagnosis lives every day — there's almost nothing.
52% Adherence Taught Me More Than 100%
I've tracked my mood daily for months. My adherence rate? About 52%. By most app standards, that's a failure. But the days I didn't log are some of the most informative data I have.
Key Insight
Logged days skew toward stability; gaps skew toward difficulty. Your dataset systematically overrepresents good days, creating survivorship bias in your own health record. You end up looking more stable than you actually are.
Reflection Days Beat Productive Days in the Data
I used to think I knew myself pretty well. Years in therapy. I journal. Then I started tracking with actual numbers, consistently. The data told me things I had completely wrong.
Key Insight
Sleep duration to next-day mood correlation was surprisingly weak (~0.3). But sleep quality to energy was strong, and energy to mood was ~0.68. Raw hours in bed was the wrong metric entirely. Also: days tagged 'reflection' averaged higher mood than days tagged 'productive.'
Breaking Your Streak Means You Needed Help
Finch has a virtual bird. Headspace gives you run streaks. Daylio has badges. Most mental health apps gamify to keep you coming back. I chose not to.
Key Insight
With a mood disorder, inconsistency in logging is a symptom, not a character flaw. The person who stops logging for a week might be in an episode. A streak counter that resets adds guilt on top of depression, making the gap longer and losing data precisely when it matters most.
I Quit My Job Because the Data Said To
Every job has stress. But if you have a mood disorder, the line between stress and destabilization is closer than you think — crossing it can destabilize your entire mental health for months.
Key Insight
'Is this stress about something or about everything?' Stress about something has an end. Pervasive dread that doesn't attach to any single cause means the environment itself needs to change. Within one week of leaving, sleep normalized. Within two, mood reversed.
Patchy Data Is Worse Than No Data
eMoods is probably the most well-known tracker for bipolar. I used it for most of 2024. It's not bad. But there's a gap between 'does its job' and 'actually helps.'
Key Insight
UI friction isn't petty — it creates inconsistent logging. And patchy data is sometimes worse than no data because it creates false patterns that point you in the wrong direction.
Your Raw Journals Never Leave Your Phone
Mental health data isn't ordinary data. It's the log you wrote at 2 AM when you couldn't sleep. The mood score that tracked a depressive episode. If any of that leaked, the damage would be personal.
Key Insight
The AI receives only anonymized, aggregated data — mood averages, stability scores, and extracted themes — never your raw journal text. Health Connect data never leaves the device at all.
Your Report Matters More Than Your Logging
If you're searching for a bipolar tracker because you want something to bring to your psychiatrist, you're asking the right question. Most mood apps don't think about the doctor appointment at all.
Key Insight
The quality of daily logging matters less than what the app does with it before your appointment. A one-page structured summary changed the entire psychiatrist conversation — she scanned it in 90 seconds and skipped straight to clinical discussion.
Source: NIMH Life Chart Method — validated for 20+ years; depression self-ratings correlate with clinician assessments at r = −0.72 to −0.79
Mood Alone Is Basically Useless Data
Most people start tracking because their psychiatrist suggested it, or because they want to understand why last month fell apart. The problem is most advice is either too vague or too clinical to stick.
Key Insight
The goal of tracking isn't to document how you feel — it's to find relationships between things. Mood on its own is almost useless. If your tracking can't answer delayed-correlation questions, it's just journaling.
Built This App at My Lowest Point
I didn't set out to build an app. I set out to not fall apart. The app was a side effect. I downloaded a bunch of apps and remember staring at five smiley faces at midnight, genuinely struggling, thinking: this is it?
Key Insight
Building during a difficult period made it better. When your mood is actually fluctuating, you catch design problems you'd never notice otherwise — like realizing a streak counter shown to someone who broke a streak because of depression is terrible UX.
Competing With Calm as One Developer
There's a specific kind of hubris that comes with deciding to compete with Calm and Headspace as a solo developer. I have it. This is what I've learned.
Key Insight
Users who pay $19/month compare you to Calm, Notion, and Linear — not to other indie apps. Every screen matters, including settings, error states, and empty states.
Your Data Is Noise Pretending to Be Signal
Search 'mood tracking app for bipolar disorder' and you get a wall of listicles. If you've tried a few while managing bipolar, you know how misleading that sameness is.
Key Insight
A mood score of 6/10 could mean a solid, stable Tuesday — or a mixed state with high energy, low stability, and declining sleep that needs immediate attention. If your app can't distinguish between them, your data is noise pretending to be signal.
Source: Sleep variability research — predicts episodes 48-96 hours before subjective symptoms appear
Track what actually matters
Mood, energy, sleep, irritability, stability — not just one number.
Get Steadyline