500 days ago, I saw a video of a woman in her 30s talking about having used dating apps for a decade, and still struggling to make a genuine connection, and noticed modern loneliness in Beijing and Singapore, in myself and friends. You may have read the three related essays [1].

311 days ago, I wrote the first line of code to a product I have been envisioning that will solve the problem.

276 days ago, Lemon was launched in Singapore. Since then, it hit 600+ signups, 50+ IRL moments curated and was featured on Singapore National TV and The Business Times.

2 days ago, I had it shut down.

Here are the 4 main takeaways from my first 0 – 1 startup and why I ultimately shut it down [2].

1. Moat

If someone can steal your idea by simply hearing it, you don’t have a moat, something defensible. E.g., “social network for pet owners” vs “low-cost way to launch objects into space” [3]

The biggest reason behind shutting down Lemon came from the lack of a real moat. When I designed Lemon, I put a lot of thought into every little detail. The end product was simple: matching compatible strangers to do interactive activities. Unfortunately, this form of IRL social product simply does not have a real moat (specifically in the context of Singapore and East Asia! I do know that Timeleft and 222 have found success in western countries).

This was particularly evident towards the end of 2025, where the trend of IRL was picking up, and everyone was spinning up their own IRL popups/interest meet ups. I have been consistently checking out other IRL social events in the scene, and even established companies that have been in the scene for 4, 5 years like 222 and Timeleft are not that differentiated from newly founded projects the likes of Lemon. You’d get a pretty consistently decent social experience at any of them if you engage actively (which most users do at these events). There is simply no real reason why any user would consistently pick Lemon over any other thing they could do, and that was reflected in the signup number. After the initial novelty wore off, in the last two months, signups completely tanked.

One can curate and provide an above average experience by increasing the burden in logistics with hosts and social facilitators that help with the experiences. However, this is obviously a great limiting factor that prevents you from scaling. You could curate a great social event maybe once a week, but struggle to replicate and scale in a sustainable way with a small team.

Takeaway:

• Consumer IRL social works as a small business, but not a startup (defined by exponential growth).
• In consumer IRL social, there is no real moat besides culture (which is inherently difficult to produce consistently).

 

2. Product thinking & real vs fake needs

There are many ways to approach designing a product. Here are some common ways I have heard/read about:

• Technology-driven: Look at what was previously not possible but is now possible due to the advent of technology (AI, VR, blockchain, etc.)
• Problem-driven: What problem or opportunities do you (in a special position) see? What is the minimal set of features that would solve the problem?
• Market-driven: What is working in one market but does not exist in another yet? (Geographic/demographic arbitrage)
• YC way: 1. Something the founders themselves want, 2. Something they can build, 3. Something few others realize is worth doing.

‘Build something people want’ is something most startup playbooks tell you, but a critical detail is often left undiscussed: what needs are real and what needs are fake, and how to tell them apart.

In Lemon’s case, it started from noticing modern loneliness as a general problem among people living in cities.  Lemon had a strong narrative: genuine IRL connections, AI recommender system for compatibility matching, the grand mission to solve modern loneliness and a pretty decent founding team (on paper).

The problem with having a strong narrative is that people love the story too much. Everyone who heard it said some version of ‘This is so needed. People are so lonely. It’s a great solution to a real problem. I’d definitely use this.’

Spoiler: they didn’t. It had almost become politically correct for people to say that they enjoy meeting people IRL and prefer ‘authentic genuine in-person connections’. Sometimes, they even do believe it themselves and genuinely show up to use your product and attend a few events. But ultimately, they all stop and return to social media and dating apps as the primary means of fulfilling their social needs.

This is one of the most important things I learned building and scaling Lemon: the difference between stated preference and revealed preference.

Stated preference is what people say they want. ‘I want to meet new people authentically.’, ‘I like to attend IRL events.’, ‘I’d definitely show up to something like this.’

Revealed preference is what people actually do. They like your social media posts but never sign up. They say they’re lonely but never sign up. They tell you they love the idea and had a great time at their first event, but never sign up again. The bigger revealed preference trend I noticed in consumer IRL social specifically is that people always go back to virtual consumer social product such as dating apps because as much as meeting people IRL without any expectations makes sense, it feels better and easier to be able to construct expectations and make judgement online through pictures and profiles and then have them fulfilled.

I had built Lemon for stated preference, but the market runs on revealed preference. Lemon didn’t fail because the problem it was trying to solve isn’t real, but because the solution required people to act against their own inertia and human nature, and most won’t.

Takeaway:

• Be careful of stated preferences, especially when people are culturally incentivized to say/act in certain ways.
• A product built around stated preferences will have poor traction because the product isn’t taking advantage of inertia of human nature and solving a real need (a revealed preference). A product that works in an ideal scenario is a product that never works at all.

3. Doing real work

Another common advice given is that you should do work that doesn’t scale, originating from and popularized by Paul Graham’s essay Do Things that Don’t Scale. The idea is that founders have to do labor-intensive, unscalable work early on, such as recruiting users one by one, providing excellent customer service, and sometimes being the product itself.

Examples of doing things that don’t scale:

• Airbnb — went door to door in New York recruiting hosts and helping them take professional photos of their listings
• Stripe — ‘Collison installation’: when someone agreed to try Stripe, the founders said ‘give me your laptop’ and set them up on the spot rather than sending a link
• Pinterest — Ben Silbermann recruited users at a design conference after noticing early users were interested in design.

In my opinion, doing things that don’t scale is so that you can achieve two outcomes:

1. Make your product better
2. Inject delight into your first users so that they feel using your product was one of the best choices they have ever made, which helps with distribution.

The core idea is that you do direct, labor-intensive engagement that builds feedback loops and compounds into growth for your product, ultimately allowing it to scale.

While ‘do things that don’t scale’ is solid advice, sometimes you might be just hustling for the sake of hustling—doing work that doesn’t matter. The on-the-ground work feels good, you feel so productive, and you have so much to signal and talk about when people ask you what you have done. E.g., “I’ve been running on the ground doing X”.

In the case of Lemon, where everyone on the team was a first-time founder doing our first serious project, oftentimes we had no idea what we were doing and just did things for the sake of doing something, so that we could pat ourselves on the back and be proud of being an entrepreneur doing entrepreneur things.

Unfortunately, signaling that work got done may be a real skill to have when working in corporate (where things always get done by systematic processes, so it is more important to signal that you got work done than it is to do actual work). In startups, nothing will move if nobody is there to push it.

Unlike school where you can get a passing grade for the effort you put in, in building startups, nobody cares about the effort you put in [4]. Your product either solves a user’s real needs and it’s a 100, or it doesn’t and it’s a 0. The danger of doing fake hustle work in activity theater is that you’re wasting your time doing things that won’t improve your product or help with distribution.

Takeaway:

• When you are doing things, ask yourself: what does this work actually do? is this real work that will help with product or distribution? If your work can’t be categorized into either category, you’re probably doing useless work.
• If your team is stuck in fake work patterns, name it early.

4. Choosing your founding team

The best founding team is one where your first project together failed, and you would still start the second one with them immediately.

One of my biggest mistakes was probably not implementing a serious and clearly defined work trial for the founding members I onboarded to the project. However glamorous/impressive their past titles, education, or accolades may be, a clearly defined work trial will show you if the person is actually capable/compatible.

Thankfully, the 1 year cliff and 4 year vesting schedule set up when onboarding new members saved me from ugly disputes towards the end of the project.

Takeaway:

• Always implement a work trial
• Always implement some kind of cliff-vesting schedule, and make clear the ownership of the project
• Experience of building 0 – 1 >> Any corporate experience/school

All that being said, Lemon was an incredible journey that I look back upon proudly. It hurts to shut my first ever 0 – 1 project down, but I am so proud of the work that was done and how far we got with it. The countless lessons on product and distribution will carry forward valuably into my future projects. Human loneliness remains a problem space that I care about deeply, and consumer social will continue to be one of my main interests.

For now, goodbye Lemon. On to the next!

Notes

[1] Four conditions for genuine connections and relationships, Modern loneliness, Rekindling human connection.

[2] Disclaimer: Everything discussed in this essay is in the context of Singapore.

[3] Eric Jorgenson, Anthology of Balaji.

[4] Andreessen Horowitz, Nobody cares.