// the take
LLRT is AWS Labs' experimental JavaScript runtime for Lambda, built in Rust on top of QuickJS. It trades JIT compilation for fast cold starts — the tradeoff is intentional and documented. Aimed squarely at serverless functions doing I/O and AWS SDK calls, not compute-heavy workloads.
The no-JIT design is the right call for short-lived Lambda functions where V8's warmup overhead costs more than it saves. Bundling the AWS SDK v3 clients natively into the binary is a real win — it eliminates the largest source of cold-start latency in typical Lambda JS code. The three bundle variants (no-sdk, std-sdk, full-sdk) let you pick binary size vs coverage instead of shipping everything. Multiple crypto and TLS backend options via Cargo features is thoughtful — you can swap in AWS-LC for optimized performance on Graviton without forking.
QuickJS has a known performance ceiling on CPU-bound work, and the README is honest that Monte Carlo simulations or tight loops are going to be noticeably slower than Node.js — but most teams won't discover this until they're in production. The Node.js compatibility matrix is riddled with partial support markers: no http/https modules natively, no worker_threads, no vm, no readline. Code that works in Node will silently break or need non-obvious bundler workarounds. It's still marked experimental, which means AWS can and does ship breaking changes between releases with no deprecation window. TypeScript is supported only via pre-transpilation, which adds a build step that's easy to misconfigure when targeting the browser platform instead of node.
We read the README, the directory tree and a key source file in June 2026. Nobody pays to be listed. How we judge →