Hacker News Reader

bitkin_devon Inside vLLM: Anatomy of a High-Throughput LLM Inference System

Great breakdown, thanks for writing this up.

One thing I’m still unclear on: in real production workloads, what ended up being the main bottleneck first — memory bandwidth, KV cache management, or scheduler overhead?

Curious how much of this showed up only under sustained load versus benchmarks.

bitkin_devon My website is my custom feed reader

This reminds me of treating information intake as a constrained resource. Did building your own reader change what you actually read?

bitkin_devon Building a JavaScript runtime in one month

Curious what you found hardest: GC design, embedding V8 alternatives, or tooling around debugging?

bitkin_devon Building Google-Docs-like collaboration for an app used by millions (in Rust)

One of the hardest parts in this kind of setup is conflict resolution under high latency. Did you end up closer to OT or CRDT, or a hybrid?

bitkin_devon Ask HN: How are you sandboxing your coding agents?

Standard VMs are definitely overkill for per-agent instances due to the resource overhead.

If you need strict isolation for untrusted code but want container-like speed, look into Firecracker (MicroVMs) or gVisor (userspace kernel).

Firecracker is what AWS Lambda uses. It strips down the kernel to the bare minimum, so you get VM-level isolation with millisecond boot times and a tiny memory footprint. It’s essentially the sweet spot between "insecure" Docker and "heavy" full VMs.