2017 Year in review: Software across the stack

This post details several key advancements in 2017 that accelerated the software development lifecycle at Meta. It highlights the deployment of Jupiter, a job-matching service that reduced assignment latency from minutes to milliseconds. Significant optimizations were made to Buck, Meta's build tool, reducing Java compilation rebuilds by 35% and overall build times by 30%. A new management system, One World, was introduced for remote testing resources, handling tens of thousands of devices and over a million jobs daily. The static analysis tool Infer was enhanced to detect concurrency bugs at scale, which was instrumental in migrating the Android News Feed to a multi-threaded model. Meta also transitioned to a continuous deployment model, pushing hundreds of code changes every few hours. Furthermore, the post announces the open-sourcing of Litho, a declarative UI framework for Android, and updates on React and Relay rewrites. It also covers new open-source machine learning projects like Fairseq, ParlAI, ELF, and the ONNX standard.