live
- your daily tech & design digest, one email a day -
Jev 1.14 adds image support and longer context [Xtra!]Incremental but steady progress on a fast-shipping agentic coding model line.
Allen AI's HiRO-ACEv2 simulates extreme weather cheaply [Xtra!]Shows ML emulators replacing expensive physics simulation at useful resolution.
Google's Deep Review tackles long technical documents [Xtra!]A concrete use case for LLMs doing deep, structured analysis rather than chat.
Routers are becoming a core AI infrastructure primitive [Xtra!]Model routing is emerging as its own layer of the stack, not just a cost hack.
vLLM ships push-based KV transfer for disaggregated inference [Xtra!]Directly cuts time-to-first-token and per-token latency for anyone running disaggregated vLLM deployments.
Netflix details how it keeps massive GPU jobs running reliably [Xtra!]Practical playbook for operating large, failure-prone GPU training/inference fleets.