Read on Twitter
Yes, you can now use the same set of tools to build your application code for both Lambda and any of your other compute platforms. No, this does not mean the existing Lambda model (managed runtimes, layers) or toolchain goes away - you like it, you keep it :).
This is not a feature for the agnostic developer. This is for those who have built depth and expertise in a docker-like toolchain, or have company wide investments in doing so, and struggle to learn something new just to get the Lambda "magic" and integrations.
Yes, you can use Lambda managed base images (which will be patched and updated in sync with our managed runtimes). If you bring your own base images, you own maintenance - just like you do with custom runtimes.
and yes, this will allow you to use the "run the same image locally and in Lambda" semantic that you like from the container world if thats your jam. Expect local testing and development to now be as easy as anything else.
One fun note - we now support 10 GB images, but without the proportional cold start impact. This the first step in us starting to do size based cold start optimizations - lots more to come here :)
oh, and one more thing - Lambda now supports 10GB/6 vCPUs as compute power. Thats our #serverless day - use your favorite toolchain, run your code with upto 6 cores per invoke, billed at 1ms. #HappyReinvent :) https://aws.amazon.com/blogs/aws/new-for-aws-lambda-functions-with-up-to-10-gb-of-memory-and-6-vcpus/
