My Tarpit Idea: Munch, that App to discover new things.

If you Google my name, you will probably find "Munch—a food discovery app" among the top results which I'm associated with. A project that I worked on with Earnest Lim, when we were students at Singapore Management University (SMU).

I decided to write this when I chanced upon a video talking about Tarpit Ideas and, hopefully, to warn off any weary travelers. What are Tarpit Ideas? Well, I will get to that later.

What is Munch?

Munch was a food discovery app co-founded by Fuxing Loh (me) and Earnest Lim, designed to address the common problem of deciding where to eat in Singapore. The app leverages artificial intelligence to recommend dining options tailored to users’ preferences, aiming to streamline the process of finding good food spots. It differentiates itself from other food-tech companies by focusing on the discovery aspect rather than just reservations or deliveries.

The development of Munch began while we were students at Singapore Management University (SMU). Incubated by SMU’s entrepreneurial program, the app received a grant and mentorship support. In total, we secured about USD 50,000 in grants and subsidies to develop the app.

Munch aims to create a more engaging and personalized food discovery experience, moving beyond simple lists to offer a robust, AI-driven recommendation system. The app was released in 2018 and was available on the Web, iOS and Android.

What did we build?

The remembrances of Munch were released into the public domain at https://github.com/munch-app

flowchart TD IOS["iOS App (Swift)"] --> CORE(Munch API Service) ANDROID["Android App (Flutter)"] --> CORE WEB["Web App (Vue)"] --> CORE CORE --> P["Partner Service"] CORE --> ES["Elasticsearch"] <--Index & Score--> D["Corpus"] AC["Article Crawler"] -- Index --> D ML["Image Recognition"] -- ML Categorize --> D S["Structured API Lambda"] -- Index --> D

A glorified CRUD app with a recommendation engine that was powered by Elasticsearch. Add buzzwords like AI, ML to index and train the data corpus, and you have Munch.

Munch iOS

The Tarpit Realization

I never quite found the words to describe Munch, and the impossibility of the idea until I chance upon this y-combinator video: Avoid These Tempting Startup Ideas

What are Tarpit Ideas?

Tarpit ideas are business ideas that seem attractive at first but are difficult to execute because they are complex, ambitious, or unwanted. They can be consumer startup ideas that seem promising but have low chances of success. Tarpit ideas can be alluring because of optimism bias, low upfront costs, and familiarity with consumer apps.

What's wrong with Munch?

What's wrong with Munch or what's wrong with discovery apps?

The magical place doesn't exist, there is a finite number of restaurants that are open tonight, and that's it. You wanting there to be a better option doesn't mean that a better option exists. The world seems limitless, but for these physical things, it's actually fairly limited.

Everyday people go on Yelp and search the restaurants in their neighborhood, and don't like what they find and are frustrated, but that doesn't mean that there are better restaurants that they don't know about. It just means that what exists is not sufficient.

Tarpit Ideas, Y-Combinator

Did I regret building Munch?

Many times, I have asked myself if I regret building Munch. Do I regret being associated with a food discovery app that sounds "so lame" in the grand scheme of cooler things I could have built? Sometimes, yeah, not going to lie. But the reality is that without Munch as that pivotal first step, I wouldn't have known what I know today and be who or where I am today.

The engineering foundation I've acquired, the lessons I've learned, and my attitude toward life could never have been without Munch. A lifetime of experience that even an MBA couldn't have given me.

Whenever I hear someone say, "I want to build a tarpit idea," I say, "Go ahead, build it." You will definitely fail and waste a lot of your time, but the version of you that comes out of it will be humbled by the version that never was.