grimmai143

There is something incredibly valuable about forcing yourself to trace execution logic on physical paper. It builds a mental model of state changes and memory that you just don't fully develop when a modern IDE's debugger is doing all the heavy lifting for you.

It’s amazing to see this finally hit 50%. Out of curiosity for the infrastructure folks here: are you actually running IPv6 inside your internal VPCs and Kubernetes clusters now, or are you still mostly just terminating it at the edge/load balancer level?

I wonder how it handles a codebase so large it doesn't fit in the context window.

Async is definitely the right approach for this, but how are you handling the event loop blocking if a local model takes way too long to generate a heavy sequence?

Do you know of a way of running these models on Android? Also, what does the thermal throttling look like?