What's your Side Project/startup regret+ Lessons
I'd love to hear any decisions people made building side projects that you learned the hard way was a bad choice
I'll start:
Deploying ML models for a project is a pain. ML models are often 100-200 MB so you cant just deploy on a lightweight VM and it exceeds free tiers. I didn't want to pay to deploy a side project that has no users.
I tried digital ocean (i got a free 300$ credit) and it was a PAIN. I spent days šgetting it working to no success. Eventually I used railway.app and it worked really well. Highly recommend. I wish I didn't try and deploy on a VM or use digital ocean for ML. Railway worked because it seamlessly downloads all the libraries from the code like heroku but cheaper but digital ocean's version of this had so many limitations with memory and ram and storage
Discussion
(2 comments)- 32 months agoTech Lead/Manager at Meta, Pinterest, Kosei
Most of my mistakes around side projects came from the idea being too complicated. The most common example of this is where the value of the app is derived from other people being on the app.
"An app to send all your friends video emojis" sounds cool, but you need to be very deliberate about how you get tightly connected groups of friends to download the app together.
Another example that is top of mind for me is promotions.fyi, which I literally just wrote about here: https://www.linkedin.com/feed/update/urn:li:activity:7212201952963674113/. The mistake we made here is that we shouldn't have even embarked on this "side project" when we had something working with Taro!
- 0
2 months agoTech Lead @ Robinhood, Meta, Course HeroUsing Firebase.
For a startup/side project, you need to move fast but that doesn't mean you should give up control entirely. Firebase has so little querying power and black-boxes so many things that you very rapidly run into concrete user asks that are literally impossible due to your tech stack.
To add insult to injury, Firebase is also comparatively expensive compared to other cloud providers and quite slow. Its performance is terrible.
If you're doing something full-stack, just do it traditionally (API layer + relational DB) or use Supabase at least.
