Loading event
Loading event

Today 10:30 – 10:50
Slush pass holders only. Coding agents can resend close to a million tokens of context on every turn, and almost all of it has been read before. This talk follows one agent request through Verda Inference, from the API gateway to the GPU. It shows how we keep long sessions cached across GPU memory, pooled RAM and SSD, and why we split reading the prompt and writing the answer onto separate servers. For developers, CTOs and anyone building with AI agents. Join us at the Verda booth (7B1) for this 20-minute talk and bring your questions. To learn more about Verda: verda.com (verda.com ↗)
From Slush, checked 12 minutes agoReport a problem

Business & networkingPeople and Culture: From a Helsinki garage to 300+ peopleToday 13:30
Business & networkingYour AI credits expire. Then what?Today 15:00
Business & networkingNobody Builds Security Alone: People, Speed, and Trust at an AI CloudToday 16:30
Business & networkingPeople and Culture: From a Helsinki garage to 300+ peopleTomorrow 11:00
Business & networkingNobody Builds Security Alone: People, Speed, and Trust at an AI CloudTomorrow 12:15
Business & networkingVerda Inference: How we serve million-token agentsTomorrow 13:45
Business & networkingRecursive self-improvement as accelerated optimizationTomorrow 15:15