PM in Progress: Reading the Tea Leaves of an App in Hot Water
PM in Progress: Reading the Tea Leaves of VibeCoding Gone Wrong
In product management, every feature choice leaves a mark. Sometimes, it’s a mark of innovation while other times, it can be a scar left on the trust of customers. Often, the key difference lies in how it was implemented.
The Tea dating advice app aimed to decode modern dating vibes and provide a social safety net for dating apps, built on the buzzy concept of vibecoding. But the team that vibecoded these features into existence failed to roll them out without proper consideration, testing, and safeguards. The result? An app that spilled user data out into the open and became a cautionary tale on how unchecked product instincts can hurt real people.
Here’s what happened and how PMs can learn from it.

How Tea Got Hot
The premise behind Tea made it go viral:
- Let users upload screenshots of dating profiles or chats.
- Offer community takes on profiles to help users mark down “red flags”.
- Provide advice and conversation starters that were meant to give users a second opinion.
As of July 2025, Tea had over 4 million female users across its platform, surging in growth by 525% from previous months and was reportedly the most downloaded free app in the U.S, ranking #1 in the app store lifestyle category.
In theory, Tea tapped into something real: that dating is messy and ambiguous, and sometimes a the opinion of another can help. In practice, vibecoding became the product’s north star, pushing the team to ship features that prioritized viral engagement over privacy, consent, and ethics. The focus became on decoding vibes, not safeguarding data.
When Tea Turned Toxic
Vibecoding was marketed as a way to “build the app of your dreams without needing to know how to code”. The vibecoded features that were designed in the direction of hype caused real harm:
- Consent Violations
- Anyone could upload another person’s dating profile or chat messages without permission.
- Within the first week, hundreds of private screenshots appeared on Tea’s servers, often without the knowledge of those being discussed.
- This risk became a reality months later when a breach exposes 72,000 images, including 13,000 selfies and IDs used for verification, highlighting just how unprotected the user data was.
2. Public Exposure Instead of Advice
- Private details with photos, names, and messages were shared openly, stripped of context and left for strangers to judge.
- A second breach revealed access to over 1.1 million private direct messages, with many containing deeply personal content such as abortion, cheating, and phone numbers.
- Tea quickly became a platform for mockery instead of helpful insights with many posts turning into public roast sessions.
3. Bad Actors Exploiting Features
- Some users intentionally uploaded profiles of ex-partners or strangers to “expose” and humiliate them. Harassment campaigns skyrocketed and became a norm for the app.
- External sites even scraped Tea’s leaked data to publicly rate women or map where they lived, fueling targeted harassment.
4. Toxic Feedback Loops
- Tea’s algorithm boosted posts with higher engagement, which usually means the more dramatic and negative comments gained more visibility and amplified the damage..
Lessons for PMs
- Trends Shouldn’t Drive Core Features
- In chasing hype, Tea lacked a foundation of ethics and safety. PMs and engineers need to ask not just “can we build this?” but “should we?”
2. Data Ethics First and Always
- Features that involve someone else’s data must have strict upload consent, moderation, and anonymization tools on day one.
3. Design for Good, Prepare for Bad Actors
- Bad Actors will exploit your feature if you give them the space to. Build controls, flags, and friction to prevent harm, not after the fallout.
4. Viral Does Not Mean Value
- Tea initially saw a spike in daily active users(DAU) as its “red flag reviews” went viral. But the breaches and hostile environment caused user trust to collapse. Long term impact was catastrophic: exposed data, lawsuits, regulators stepping in, and a once promising app became synonymous with harm instead of safety.
- Sustainable growth comes from user safety and positive experience, not sensationalized engagement.
The Final Sip
Tea wanted to make dating safer. Instead, vibecoding turned it into a gossip machine that leaked user data and eroded trust. The cautionary tale is that when hype leads feature design, real people get hurt. PMs should ask:
- Does this feature truly protect and serve?
- How could it be misused and do we have the safeguards in place?
- Would I feel safe if this feature were used on me?
Because when it comes to rolling out features, once it spills, it’s hard to clean up the mess.
About Me
Hi, I’m Andrew! I’m a student, product management enthusiast, and feature detective. PM in Progress is my corner of the internet where I break down what works in tech products that we use.
This post came from my curiosity about how hype can turn a “cool feature” into a cautionary tale. My goal is to keep learning how we can ship products and ideas that people can trust!