Engineering insights on API reliability, distributed systems, and making failure boring.
By 2029, AI agents and embedded devices will become too chaotic without a new internet protocol. No company is building the bridge from serverless to operationless. At least until I built EZThrottle.
Read more →Announcing AccountQueues: per-user, per-key queue isolation that solves the noisy neighbor problem. Every customer gets their own lane. Plus: two response headers that let API providers control their own flow rate, and why our 2 RPS default protects the internet.
Read more →US-EAST-1 went down. Somebody got fired. The worst outages aren't dramatic — they're invisible. Your workflow isn't crashing, it's just slow. Here's how to automatically reroute around regional failures.
Read more →How to scale AI agents without burning out engineers. LangGraph handles state beautifully, but when OpenRouter returns 429 at step 7, your workflow crashes. Here's how to fix that with coordinated retries, fallback racing, and webhook-based resumption.
Read more →What happens when you stop managing retries and start shipping features. Replace Lambda, SQS, DynamoDB, and Step Functions with one API call. Scale requests, not pods. Build in a weekend what used to take months of IAM policies and debugging.
Read more →Modern software fails with 429s, timeouts, and regions going dark. I built EZThrottle because I got tired of pretending this was normal. By turning retries, rate limits, and region health into shared state, it gives you something rare in distributed systems: predictable behavior when things go wrong.
Read more →Not accepting public signups. Check out Aquifer, the open-source runtime this is built on.
See Aquifer on GitHub →Want a private cluster? support@ezthrottle.network