Jev 工程师做流量镜像测试终端，做成月入 1.8 万美元服务
把线上请求镜像到隔离环境，新代码影子运行，Jev 逐请求比对行为而非文本，失败点可追溯到具体组件，再决定是否放量。
A RANDOM JEV ENGINEER BUILT AN AI TERMINAL THAT REPLACED HOURS OF CI TESTING AND TURNED IT INTO AN $18,000/MONTH BUSINESS. Instead of waiting for pipelines to finish, he built a visual system that tests changes against real traffic before they ever reach users. The trick is that the new code never touches the real response. > MIRROR - live requests copied into an isolated testing environment > SHADOW - new code runs beside production without affecting users > JUDGE - both outputs compared request by request, not by averages > TRACE - failed checks mapped to the exact component that caused them > SWAP - verified changes promoted with a rollback ready Traditional CI pipelines test against expected behavior. Real traffic exposes the cases nobody thought to write tests for. This terminal makes that difference visible. Every request becomes a path through the system. Failed checks light up, dependencies connect, and the exact point of failure becomes traceable. The comparison is where things get interesting. Two responses can look different and still be correct, so the system needs to judge behavior rather than exact text. The biggest advantage isn't speed alone. It's catching a bad change before a real user ever receives it. And the business model is straightforward: turn the setup into a repeatable deployment-safety service for engineering teams. The expensive part was never waiting for CI. It was shipping the one bug CI failed to catch. The full workflow, from mirrored traffic to automated rollback, is in the breakdown below.
这条是引用着另一条写的：x.com/0x_mura/status/2099183948941115418