Loom already delivered virtual threads and scoped values. It still needs to deliver the Structured Concurrency API, the element that you need to launch network requests in parallel. As of now, launching requets in parallel can be done using the ExecutorService pattern, or callback-based asynchronous programming. The Structured Concurrency API gives you new patterns of code, based on virtual threads. The code you write is more readable, with less places for bugs to hide. It also gives you meaningful stack traces for better debugability and profiling. This presentation takes you through these patterns, showing you how the StructuredTaskScope objects are working, how you can configure them. You will see how you can build your own asynchronous strategies for the handling of the results your tasks give you using the Joiners that the API gives you, and how you can build your own Joiners for the specific needs of your application.

Jfokus

A conference talk by José Paumard (Oracle Java Developer Advocate) covering Java's Structured Concurrency API, still a preview feature in JDK 25/26. The talk explains why traditional ExecutorService patterns are problematic (poor stack traces, thread leaks, security gaps with ThreadLocals), and how Project Loom's StructuredTaskScope fixes these issues. Key concepts covered include: creating a StructuredTaskScope via the open() factory method, forking tasks that run on virtual threads, calling join() before retrieving results, and using the Joiner interface to customize behavior on task completion, failure, and timeout. The talk also explains how ScopeValues integrate with structured concurrency to safely share bindings across virtual threads without escape risks. Practical warnings are given about race conditions in the onComplete callback and the importance of not using virtual threads for CPU-bound work.

Make Your Code Parallel with the Structured Concurrency API, José Paumard