Nov 2 · Remi · I filed Remi's trademark applications with IMPI, Mexico's trademark office. ›
← Blog

SpeakOn: the app that hit #1 on the App Store in Mexico, and why it died

SpeakOn was my first big tech project: a messaging app that translated in real time across more than 60 languages. In June 2015 it hit #1 among free apps on the App Store in Mexico, ahead of WhatsApp and Facebook. And it didn't survive, because it was built so badly it couldn't handle its own users.

Carlos Gotlib5 min read
Carlos Gotlib in a navy blazer looks at his phone next to the text It hit #1. And died.
In short
  • SpeakOn got 200,000 users in 48 hours and 350,000 sign-ups in total.
  • It died because the tech couldn't handle the load, and fixing it meant rebuilding it from scratch.
  • The lesson: if you don't understand how your product is built, you don't control your company.

SpeakOn hit #1 among free apps on the App Store in Mexico in June 2015, ahead of WhatsApp and Facebook, with 200,000 users in 48 hours. And it died from that same success: the app was built for a small number of users, it crashed exactly when the most people wanted in, and fixing it meant rebuilding it from scratch.

It was my first big tech project. Here I explain where it came from, what launch day was like, why it crashed and what changed in me afterward.

Where did the idea for SpeakOn come from?

From a problem of my own. Back then I talked a lot with suppliers on Alibaba and we couldn't understand each other. I'd write in Spanish or English, they'd reply however they could, and half the conversation got lost.

I imagined a chat where everyone wrote in their own language and the other person read it in theirs. That was SpeakOn: a messaging app that translated in real time in 60 languages. A normal conversation, like in any messaging app, but without the language barrier.

Who built it?

I didn't code. I posted jobs on Indeed and elsewhere, and I also hired people from UNAM who were supposedly very good. I ended up with a team of four developers: two backend, one Android and one iOS.

I didn't look for a technical cofounder. I didn't think I needed one. I figured that if I hired good programmers, everything else would sort itself out.

The backend was built in Go (Golang) with MongoDB as the database.

Launch day

We threw a launch party at Hotel Habitat in Mexico City, with celebrities and press. It went viral.

At the office we had a screen showing downloads and sign-ups in real time. The numbers were going up roughly one per second.

In June 2015 SpeakOn was the #1 free app on the App Store in Mexico, ahead of WhatsApp, Messenger, Facebook, YouTube and Spotify.

| Metric | Figure |
|---|---|
| Users in the first 48 hours | 200,000 |
| Total sign-ups | 350,000 |
| Languages with real-time translation | 60 |
| App Store ranking in Mexico | #1 free app, June 2015 |

SpeakOn at #1 among free apps on the App Store in Mexico, above WhatsApp, Messenger, Facebook and YouTube

Screenshot of SpeakOn at #1 among free apps on the App Store in Mexico, above WhatsApp and Facebook. SpeakOn archive.

Why did SpeakOn crash?

At some point people kept downloading the app but couldn't sign up anymore. Then everything went down, including the website.

Picture the worst possible moment: thousands of people who just heard about your app download it, excited, and can't even create an account.

I looked for help and ran into another problem. The Go community back then was small and closed; it wasn't easy to find someone who understood what we had. I ended up bringing in two of the best Go engineers in the United States.

They reviewed everything and told me: "This has to be rebuilt. It can't be fixed."

They explained it with an image I still use. Building a system is like building a house. If you didn't design it to hold a pool on the top floor, adding one later is nearly impossible. You have to tear it down and build again.

What happened next

We tried to rebuild it. It took too long, and users who come for the novelty don't wait. We couldn't keep them.

Two other things happened during that stage.

Bessemer Venture Partners reached out to us because they were interested in investing.

I applied to Y Combinator. We made it to the first interview and weren't accepted. The reason was the lack of "technical skills" on the founding team. That was exactly my problem: nobody on my side understood how the product was built.

The mistakes that killed it

None of them were coding mistakes. They were my decisions:

  1. I had no one technical on my side. No cofounder, no advisor. Nobody was reviewing the team's decisions with independent judgment.
  2. We built for the launch, not for success. The architecture worked for a small app, and the launch plan was for a big one.
  3. We didn't test under real load. We never simulated what would happen with hundreds of thousands of sign-ups at the same time.
  4. We launched to everyone at once. The party, the press and the celebrities brought everyone in on the same day.

I go into detail in What I learned from outsourcing the development of my first app.

What changed in me: I learned architecture

The most important lesson from SpeakOn was this: if you don't understand how your product is built, you don't control your company.

So I decided to learn. I studied code and architecture. Today I understand architecture, infrastructure, services and microservices. I can read code, even if I don't write it.

I didn't become a programmer. I became someone who knows what to ask, understands the answer and notices when something doesn't add up. That's the difference I was missing in 2015.

Was SpeakOn a good idea?

I think so. Ten years later, real-time message translation is a built-in feature. In June 2025 Apple announced Live Translation inside Messages, FaceTime and Phone.

The demand we saw on that office screen was real. But a good idea won't save you if the product can't hold up on the day it works.

What I'd do differently today

The world changed recently. What I learned in 2015 doesn't mean I need to hire someone technical to build every app today. AI changed the tools I have to do it.

  • Separate what I needed in 2015 from what I need today. With today's AI tools, I could build something like SpeakOn without someone technical on my side. For hard tech or greater complexity, I would still need that person.
  • I'd define how many users it needs to handle and ask for the design to prove it.
  • I'd run load tests before the party, not after.
  • I'd launch in stages, with a waitlist or invites.
  • I'd understand enough to ask the right questions. Today I do.

Is a viral app a business?

Not necessarily. Hitting #1 tells you people are curious. It doesn't tell you whether your product holds up or whether anyone will pay.

Years later it happened to me again, in a different way, with Donna. That time the tech held up; what didn't work was the business model.

The full story of my companies, in order, is in who I am and why I now build in public.

Frequently asked questions

What was SpeakOn?

A messaging app with real-time translation in more than 60 languages, founded by Carlos Gotlib in 2015. Each person wrote in their own language and the other person got the message in theirs.

Did SpeakOn reach #1 on the App Store?

Yes. In June 2015 it was the #1 free app on the App Store in Mexico, ahead of WhatsApp, Messenger, Facebook, YouTube and Spotify.

How many users did SpeakOn have?

200,000 users in its first 48 hours and 350,000 sign-ups in total.

Why did SpeakOn shut down?

Because its tech couldn't handle the user load. The app was poorly built, and fixing it meant rebuilding it from scratch.

What was SpeakOn built with?

The backend was built in Go (Golang) with MongoDB, by a team of four developers: two backend, one Android and one iOS.

Who founded SpeakOn?

Carlos Gotlib, founder and investor. He later founded Donna (formerly WeHelp) and Kimik, and today he's building Remi in public.

Learn how to build a company by watching me build mine.

Every week: the decision, the number, the mistake and the win.