AETHER
Test in Production , Break Nothing .
The problem AETHER solves
I solve the fundamental crisis of modern software: 'Staging is a Lie.'
We spend billions building sterile testing environments, but they never match the chaotic reality of production. Because of this, we accept that every few months, a bad deployment will crash the site or corrupt data. Existing tools like Feature Flags or Canary Deploys are just damage control—they still use real users as 'mine detectors.'
The specific technical problem I solved is the 'Double-Write Paradox.'
Standard traffic mirroring fails because you can’t mirror a 'Buy' request without charging the customer twice. I built a Virtual Database Proxy that traps these side effects at the driver layer. This allows me to test risky code against live, dirty production data with mathematically zero risk to the end user.
I turned Reliability from a luxury into a guarantee.
Challenges I ran into
The hardest engineering hurdle was guaranteeing 'Zero-Impact Isolation.' I ran into two specific catastrophes that I had to engineer my way out of.
First was the 'Concurrency Race,' specifically Mutex Contention. Initially, I used a shared lock for logging. I discovered a fatal flaw: if the Shadow thread crashed while holding the lock, the Production thread would freeze waiting for it. I essentially created a deadlock where a test failure took down the real site. I solved this by decoupling the locking mechanisms entirely, giving Production its own lockless path so that nothing the Shadow thread does can ever block the main thread .
Second was the 'Zombie Connection' problem. When a user closed their browser tab, the backend didn't know. It kept trying to push WebSocket messages to a dead client, which filled the memory buffer and crashed the server. I had to implement a strict ping/pong heartbeat and a WriteDeadline in Go to forcibly sever these zombie connections and free the memory .
Finally, conceptually, solving the 'Double-Write Paradox' without using expensive mocks required me to build a custom Virtual Database Proxy at the driver layer to intercept SQL commands in real-time.
Tracks Applied (1)
Open Track
Technologies used
Cheer Project
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.