At the recent KVCache Meetup Korea, co-hosted by PyTorchKR and XCENA, speakers from XCENA, SK hynix, FuriosaAI, and LMCache dove deep into the architectural trade-offs powering large language model serving. Talks spanned CXL memory pooling with Maru, roofline analysis of GPU compute and HBM bandwidth during prefill and decode, batch scheduling on RNGD, and process-isolated cache serving with LMCache MP Mode. Real-world optimization requires collaboration across every tier of the stack, from accelerators and memory architectures to runtime engines. Thank you to PyTorch Foundation Ambassador Junghwan Park, co-host XCENA, and sponsor @tensormesh for bringing this local developer community together. Catch the full event recap here:

$23 billion in AI infrastructure demand. Then $99 billion the year after. That was the opening salvo from Ben Richardson (@CoreWeave) at RAISE Summit, and the number that silenced the room. Anthropic alone is projected to go from 1.5 gigawatts of compute to 10 by 2027. The equivalent of building today's Google, from scratch, in two years. Four companies racing to do the same simultaneously. @JeremieEO (@SemiAnalysis), June Paik (@FuriosaAI ), Vik Malyala (@Supermicro), @alex_saroyan (@netris_io), and Shaun O'Meara (@MirantisIT) unpacked what that actually requires, the energy, the infrastructure, the supply chain, and why none of it was built for this moment. Watch the full session: #ReliveRAISE#

- @adnansvirk 's Sports Emmy Awards Recap - Kevin Costner’s big gamble - Furiosa - The Jinx: Part 2 - The Life and Death of Colonel Blimp - Adnan sits down with director Tony Goldwin to talk about his new film Ezra.