So it’s official. didau.co, the Vietnamese version of my AI travel planning site launched on 22 May.
Well, an accidental launch. I put the site up in production mode to get some feedback from friends and one of them shared it on his Facebook page (about 60 people in total). Since then, it attracted about 5-6 users daily; I might as well just leave the site up.
What is didau.co?
In short, you put in any holiday request you have and the AI under the hood will generate a travel plan with attraction highlights, a suggested itinerary, related accommodations & transportation methods.
For each activity or hotel mentioned, the AI links to one of my affiliate partners. If a user books a product, I get a cut of that.
Very basic affiliate business, just that I crowdsource the generation of content through users & AI.
What is the result so far?
I made $2.
My first internet magic money. It was an exhilarating moment even though the money isn’t enough to cover a meal.
I made my first buck not from working for others.
Here are some stats (currency in VND)
Reflection
I was totally surprised by this.
The reason? I didn’t expect the site to make any money just yet. I didn’t plan to launch it so early; thus the site still misses several critical features in my book:
No signup/login. Not even a form to collect emails.
No way to gather feedback: no tracking installed (GA/Mixpanel/Amplitude), no feedback form to pop up if a user has stayed long enough.
No SEO setup. The site is not up for Google indexing although it’s my main strategy to generate traffic. Heck, even now, I have yet to set up a title tag or a meta description tag, or a robot.txt file.
Broken mobile layout on the home page. You can type in the search box but there’s no button to click to go to the itinerary page.
Yet, somehow, people book trips through my site. And I have no idea why they do.
I spent about 90 hours building this site. In reality, it’s more like 1.5 months with a 1-year-old baby to look after at the same time (indie hacker parents please take note - this is you playing the game at God of Sparta mode).
Now I believe I launched too late. I built features that ended up didn’t make the cut:
AI:
Generate itineraries from a blog RSS feed. Didn’t have enough data for travel locations to matter.
Passing user queries to Google, analyzing & summarizing the search results to create an itinerary. I threw it away because the speed to get the AI to write the first word is just too slow (10s).
I ended up with prompt engineering directly to the LLM and the quality is just good enough in a decent time.
Internationalization - I’m going to launch an international version at plantrips.net so I decided to support i18n from the start. However, I went a bit too overboard and try to be smart with a language switcher, saving locale in cookie & falling back on the server. This could be just an env var.
Playing around with fly.io for an auto-deployment goodie. It couldn’t handle my 8GB Python docker container image. In the end, I went back to a simple EC2 instance with dokku push to deploy.
Not doing these above would save me about 20 hours of work. Yet, I still think it’s too long for a launch. Hear me out:
An average bootstrapped app (not a moonshot VC type of idea) has a 1/100 chance of success. With a 1.5-month per app launch velocity, you theoretically could build 8 apps a year. That means it roughly takes 12.5 years before you can rely on bootstrapping as a main source of income for yourself. Not to mention supporting a family.
As a result, the only solution is to launch faster. Over the past month, I’ve seen people who launch on Twitter with just a landing page and a stripe form after a weekend of work. That will be my new target to aim for.
Future plan
I’m putting a soft deadline to launch plantrips.net by 2 June and put it up on ProductHunt the next week, which is partly the reason I write this piece. Just putting myself out there for some shame if I miss it.
Adios, till the next launch report.
Ps: Good discoveries since last month
posthog.com if you think of starting a side project and want to understand how users use your product. Their Session Replay feature is totally nut and allows you to see where users spend time/focus on your site. As an added bonus, they include error reporting in replays so I also don’t need to set up Sentry just yet :D
convertkit.com for email marketing, is much cheaper & easier to set up compared to Mailchimp
Rob Walling’s Stair Step method to software bootstrapping. An old but golden piece of strategy for indie hackers/solo founders. His portfolio of small bets approach is what I’m trying to replicate moving forward.
Yay! Congrats on a new path of life ^^