The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Spent 4 years building iOS in-app purchases from scratch. Here's the one thing I wish someone told me earlier:
Test with TestFlight, not just Sandbox.
Sandbox accounts behave differently than real purchases. I had "working" IAP that broke in production because sandbox hid edge cases.
Other quick lessons:
Receipt validation must be server-side (client-side = getting scammed)
Apple rejects vague IAP descriptions — be specific
"Restore Purchases" button isn't optional
These got me a 100% App Store approval rate across multiple submissions. Currently taking on React Native projects if anyone needs IAP help or App Store submissions 🚀
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Flagged payments don't wait for you to get back to your desk. So Halden's overview works on the phone too.
Great work, as always! 🎉
This is AI Lab Eval Platform I built for data science teams - monitors AI model performance, eval results, and failures in real time.
Problem I solved: Finding where AI systems are wrong. I designed evals to test accuracy on large complex research datasets (1.2M rows).
Key features: Accuracy tracking (87.4% +2.1% improvement), Eval Failures detection (23), Dataset Size handling, Accuracy chart over 8 experiments.
Tech I used: React + TypeScript for virtualized table (1.2M rows without lag), TanStack Virtual, Recharts, and Claude API for AI-powered query & insight.
Right panel: User asks "Show me all eval failures from last week" and Claude gives key insights - data leakage, prompt injection, low confidence outputs with recommendations.
Result: Shipped ML changes that moved accuracy from 70% to 87.4%. Built full-stack product surface that renders large datasets efficiently. Ready to build same for your AI Lab.
That looks cool!