Every few months, a Product Hunt launch crosses my feed that teaches me more about content than the latest AI scheduling tool. CoachAI was that launch. It’s an iPhone app that watches a workout, spots when form breaks down, and tells you in real time — all on-device, nothing uploaded. On the surface, it’s fitness tech. Underneath, it’s a masterclass in the exact problem every creator and social media operator is chasing: earning attention before asking for anything. The maker’s pivot from five screens of onboarding to “one tracked rep first, then ask questions” is a hook structure that belongs in every content calendar, every funnel, every pitch deck.
The Feedback Loop Is the Product
Most social media tools optimize the calendar, not the learning. I can schedule thirty posts across five platforms in an afternoon, but the tool won’t tell me whether any of them earned a second look. That’s the gap CoachAI is designed for in a completely different domain: it creates a real-time feedback loop. According to Mouhab Tarek, the maker behind CoachAI, the app uses the iPhone camera to track movement in real time and tells you when your form breaks down, all on-device with nothing uploaded. Around that sensor sits a program engine that adjusts the workout instead of shuffling random exercises.
In my own tests of AI tools for content work — auto-captioners, editing assistants, engagement predictors — the best ones share one trait: they tell you why something happened, not just what happened. CoachAI does that at a movement level: it senses, judges, and corrects. Compare that to the incumbents. Buffer and Later solved the scheduling layer years ago. Hootsuite and the analytics crowd solved the dashboard layer. They answer “when” and “how much,” not “why.” That’s why CoachAI’s mechanics feel closer to what creators need from analytics than most analytics dashboards. A content engine that watches a rough cut and says “this opener will lose people” before you publish — that’s the social media version of CoachAI. It doesn’t exist in most stacks yet. When it does, we’ll wonder how we tolerated the delay.
My take: the fitness domain is a shell. The real product is the feedback loop. For a creator, the equivalent feature is not prettier analytics. It’s a system that can tell you, before and during production, where the attention drops. CoachAI already does that for a squat. The same logic will be built into every serious content platform eventually.
Show the Magic, Earn the Form Fill
Tarek’s original onboarding asked for goals, height, weight, injuries — five screens before users ever saw the thing that made the app different. He says people bounced. So he flipped it: you do one tracked rep first, then the questions come. As he put it in his Product Hunt launch post, “Show the magic, earn the form fill.”
That’s the best content strategy sentence I’ve read this year. Most social content does the opposite. We open with the logo, the intro, the request to subscribe, and the “story of how I started” before we deliver the actual value. Then we wonder why retention drops. The content funnel mirrors the onboarding funnel: ask for commitment before proving the outcome, and you lose the people who don’t already know you.
I now use this as a test. If a post can’t deliver some value in the first five seconds, I rewrite the opening. In my own planning sessions for a client’s short-form content, the videos that opened on the finished output — the before-and-after, the final slide, the result itself — were the ones I didn’t have to defend in review. The videos that opened on context took twice as long to justify. That’s an observation, not a controlled experiment, but it maps to how the averages work: cold-start feeds do not know you and will not wait to find out why they should care.
Why TikTok creators should care more than LinkedIn ones
Not every platform needs the same opening. LinkedIn still rewards text posts that set a scene and build over a few paragraphs; a cold opener can work if the payoff is a professional insight. But TikTok’s algorithm is gnarlier: watch time, rewatch sessions, and completion rate decide distribution before follower count ever matters. If the first second doesn’t earn the second, the platform buries the video. CoachAI’s onboarding flip is a TikTok-native move: show the result, then earn the follow. LinkedIn can wait. TikTok never does.
The Angle Problem Is the Algorithm Problem
The best question in the thread came from Omri Ben-Shoham, who pointed at the camera issue. He worried that a phone propped to one side can miss depth or spinal rounding on a squat or deadlift, and that a false positive is worse than silence: if the app gives a thumbs-up when the angle didn’t have visibility, people trust it and load more weight. That’s not a nitpick. That’s the same failure mode as a social analytics dashboard that reports engagement without reporting the confidence of the measurement.
CoachAI says it’s on-device and lightweight — good for privacy. But on-device also means the model is limited by what one camera angle can see. Tarek asks for harsh feedback on tracking accuracy, which is the right way to launch a product that depends on trust. But the open question is whether the product knows what it doesn’t know. Does it tell the user when the camera angle isn’t good enough? The source doesn’t say. I’d treat that as the most important missing feature.
Where the math breaks: AI confidence without visibility
Every AI tool that gives a confidence score should also give a visibility warning. CapCut can auto-caption a video, but it won’t tell you that the audio is too muffled for the captions to be trustworthy. Instagram can suppress a post, but it won’t tell you why. The absence of feedback feels like a pass. Tools that admit “I can’t see well enough to judge” are rare, and that’s exactly what CoachAI could do differently — if it doesn’t already. I’d bet the tracking-accuracy work Tarek mentions is really an uncertainty-measurement problem in disguise. My take: a tool that articulates its blind spots is worth more than a tool that is always sure.
The Comment Section Is a Product Roadmap—But the Rough Edges Are Real
The two comments I keep thinking about are the expansion questions. Elliot Myhre asks about using the same approach for tennis and getting real-time feedback through AirPods. Ethan Cheng asks about analyzing all types of movements. Neither question is answered in the scraped thread, which is fine — the questions themselves are the signal.
For a social media operator, this is the difference between reading comments for validation and reading comments for product intelligence. The best market research you’ll ever get is people asking “can it also do X?” because X is a feature, a new audience, or a content vertical. If I were running CoachAI, those two comments would be pinned to the top of the content calendar. They are the next version.
But let’s balance the enthusiasm with the limitations. CoachAI requires iOS 17 and is free to start, according to the App Store listing, but “free to start” implies a paid path that isn’t disclosed in the source I have. Android, web, export, and API access? Not mentioned. The maker is upfront in his launch post that the app is “still early, still rough in places” and asks for harsh feedback, especially on tracking accuracy. That transparency is a trust signal, but it also means you are beta-testing a tool that could miss an injury-relevant detail.
So who is this not for? Experienced powerlifters who need barbell-path measurement probably still need a coach or a dedicated tracking system. Coaches trying to monitor remote clients need sharing and export tools that are not disclosed. Android users are out. And social media managers looking for a content scheduling tool? This is not that — don’t buy it for the dashboard. Steal the philosophy instead. My take: CoachAI is not a replacement for a human coach. It’s a pre-coach — a cheap, on-device cue that asks whether you’re ready to put your body under load. The moment it tries to be more than that, the camera-angle problem becomes a liability, not a feature.
What I’d Watch / Test Next
This week, I’m taking three concrete steps, and I’d recommend the same to any creator or operator reading this:
- Flip one piece of content. Put the result before the context. Use CoachAI’s “one tracked rep first” as the template: let the magic live in the first five seconds, then ask for the save, follow, or signup.
- Audit every analytics dashboard you trust. Ask it Omri’s question: can this tool see the angle it’s judging from? If it can’t tell you when it’s blind, treat its recommendations as suggestions, not verdicts.
- Mine the comment section for “can it do X?” questions. That question is the next feature, the next post, the next product.
And one meta-step: ship one rough thing this week. Label it beta. Ask for the harsh feedback. A rough product that reaches real people is worth more than a polished one nobody sees — that’s not just Tarek’s line. It’s the entire thesis of this launch.
