Matin Tamizi spent his between-companies years the way many founders do after an exit: training for races he probably shouldn't have signed up for. Sleep-deprived from a newborn, traveling for work, occasionally sick—and yet his workout app kept prescribing intervals as if he were living in a monastery.
The frustration stuck. Now, fifteen years after his first Y Combinator batch and several fintech exits later, Tamizi is back with Imperfect, an AI coaching platform that does something most training apps studiously avoid: assume your life is messy.
Launched publicly in early 2026, the app sits at an awkward intersection. On one side, there's a sudden stampede of fitness companies bolting AI onto existing products—WHOOP's OpenAI-powered Coach arrived last May, followed the same day by an agentic assistant from DICK'S Sporting Goods. Garmin continues refining adaptive plans. On the other, athletes are drowning in wearables data with no good way to make sense of it all when real life intervenes. Tamizi is betting the gap between those two realities is wide enough to matter.
An App That Reads the Room—and the Weather
What Imperfect actually does might sound straightforward until you realize how few platforms bother with the details. The app pulls metrics from Garmin, WHOOP, Oura, Apple Watch—standard stuff. Then it layers on the kind of context that rigid periodization plans treat as noise: air quality warnings prompt indoor alternatives, poor sleep triggers workout adjustments, early signs of illness dial back intensity before you're flat on your back.
The interface leans conversational, which in practice means interactive notifications after workouts and a voice check-in feature that lets you skip the typing. By late spring, the app supported uploading photos and PDFs—race schedules, injury notes, meal logs—and tracking full calendars of events. It syncs with weather patterns and your actual schedule, the kind of micro-adjustments that separate a plan you follow from one gathering digital dust.
Perhaps more telling: the nutrition prompts. Instead of asking you to hit abstract macro targets, Imperfect translates protein goals into food. "That's about three tacos." Small touch, but it suggests someone who's actually tried eating enough protein while stuck in airports.
The app covers running and cycling, naturally. Also jiu-jitsu, padel, MMA, snowboarding, bouldering, surfing, skyrunning, aquathlon. Whether that breadth reflects actual usage or aspirational positioning three months in remains unclear.
Infrastructure Instincts
Tamizi's previous life building financial infrastructure shows up in unexpected places. He built Balanced, which shut down in a transition deal with Stripe, and founded Cuenca. Before launching Imperfect, he authored an open-source Python library for Garmin Connect authentication—code that Stanford Health reportedly now uses, according to the company. That technical fluency extends to transparency uncommon in consumer fitness apps: Imperfect explicitly discloses using models from both Anthropic and Google to generate recommendations.
It's the kind of detail that matters less to users than to investors evaluating defensibility. Which raises questions, because the company has disclosed only standard Y Combinator funding—$500,000 total, with $125,000 for 7 percent equity and the rest on an uncapped SAFE. No follow-on rounds have been announced.
The team stood at five people as of June 2026. That's lean even by early-stage standards, particularly when you're trying to support nearly a dozen sports and integrate with multiple wearable platforms simultaneously.
Crowded Territory
The timing complicates things. Strava acquired training app Runna in April 2025, then scooped up The Breakaway in May 2025—a clear signal it wants to own coaching atop its recording and social dominance. WHOOP, fresh off a fundraise in March 2026 that valued it at $10.1 billion, is pivoting hard toward "intelligent health platform" territory with on-demand clinician access alongside its AI coach. Therabody launched an AI recovery coach at CES.
Imperfect's pitch differs more in emphasis than approach. Where WHOOP's AI leans on proprietary biometric hardware and Garmin's plans live inside a mature device ecosystem, Imperfect exists as connective tissue between platforms. The bet: athletes don't want hardware lock-in, and daily life micro-adjustments matter more than perfectly periodized training blocks.
Maybe. While purported to be available on Android, it currently lacks a Google Play listing, though the company's terms of service mention Android—suggesting either imminent expansion or language that got ahead of engineering reality. Early App Store ratings hover around 3.0 across a handful of reviews, with at least one user complaining about slow AI response times. The sample size is too small to mean much.
The Monetization Mystery
Here's what Imperfect isn't yet: a business. The app remains free, and the company stated back in April that there would "always be a free plan." What sits above that tier, when it arrives, hasn't been disclosed. Premium features? Advanced analytics? Earlier access to new sport integrations?
The silence might be strategic—no need to telegraph pricing before product-market fit is clear. Or it might reflect uncertainty about what users will actually pay for in a market where Strava already charges $80 annually and WHOOP requires buying hardware.
Tamizi has built companies before, sold them to names like Andreessen Horowitz and Stripe. That track record buys credibility. But the question isn't whether he can build products—it's whether the space between perfectly structured training plans and chaotic daily schedules is wide enough for another venture-scale company to emerge, particularly one without unique hardware or a social graph to defend.
The answer probably depends less on whether the AI is clever and more on whether athletes care enough about daily life adjustments to switch platforms. Which is another way of saying: we'll know more in a year than we do right now.
