Directory
Note: As a startup, we move fast & adapt & iterate. Assume everything here is outdated. Here are some of our previous versions:
Engineering FAQ
1. Scaling an App from 0 to Millions
I compiled a 68-page document detailing the engineering challenges of building a billion-dollar company.
2. What’s your Tech Stack?
The cardinal rule of engineering: Every decision has tradeoffs.
React Native forces you on Expo for web, which may lead to performance issues down the line. Fear of fines (associated with data breaches) may sway you to adopt a “secure” language like Java, used by enterprise customers and financial institutions alike.
Peers 1.0: React Native + Tamagui + Supabase (CDN, Authentication) + PostgreSQL (Database) + Stripe (Web Payments), RevenueCat (Mobile Payments)
MERN Stack: MongoDB, Express, React, Node
OnlyFans: Flutter, MySQL, JQuery, Node
3. Sorting the Comments Section?
An active comments section is one of our unique selling points. How do we design a killer comments section that provides value to all?
a. Most Likes Wins
The most basic option. Substack uses this as its default.
b. Ratio of Likes to Dislikes
Introduce a two-sided weighing system. Allows users to self-moderate by flagging inappropriate comments with a “downvote” feature.
c. Introduce a Time Penalty
Do we want to immortalize old comments to create an “in-the-moment” feeling? Do we want to prioritize newest comment first?
d. Integrate Noise
An anti-bot measure. Reddit injects noise into their calculations (“fuzzing”) so bots can’t tell if their likes counted.
e. Add Custom Weighting
For example, comments made by higher-tiered subscribers may count more. This is how you “shadowban” or “de-boost” someone.
f. Predict the Future
You can use a statistical measure like the Wilson Score Interval to surface the best comments of the future.
g. Optimize with Machine Learning
You can get as fancy as you like. Recently I read a paper that explored using Latent Dirichlet Allocation (LDT) to calculate word groups then sorting by Cosine Similarity and then running through a Siamese Neural Net (used by fingerprint scanners).
4. Web vs Mobile?
Web-first. Why?
Entry Cost. Downloading an app on the App Store is more difficult than navigating to a URL.
Development Cost. Mobile developers are more expensive to hire than web developers.
Cross-Compatability. A responsive website can be accessed on mobile whereas a mobile app cannot be accessed on a browser.
Speed of Iteration. The App Store or Play Store requires approval, which can take 1-3 days. Web deployments are instantaneous.