78% of the top 50 generative AI web products by traffic run on Cloudflare. Builders working on the frontier of AI didn't choose Cloudflare for a feature list. They chose a set of architectural decisions made years before agents arrived — decisions that turned out to be exactly what AI workloads need. In this 20-minute presentation, Rita Kozlov, VP of Product at Cloudflare, walks through what those decisions were and what they mean for enterprise teams adopting agents this year. Watch here: https://lnkd.in/eUzxe6sy
thrives
1mo
78% sit on Cloudflare because buyers trust architectural decisions made years before agents arrived, not the demo circuit. The interesting play is whether the next wave breaks to someone cheaper. The eval that answers that is cost-per-successful-task at production traffic, not at launch. thriv.es would gladly compare notes on the instrumentation side.
Sabhi Sharma
1mo
Really Interesting statistic, in the near future not every org would be able to accommodate parallel agentic workflows into their regular workflow. This might create an AI divide between organizations which can afford and others who can't afford.
Michael Poley
1mo
Very interesting. Near term it seems like not everyone gets access to these CPU resources, sounds like interesting times for builders.
The most interesting part is the architectural perspective behind statistic.
Justin Liu
6d
One thing I find interesting about the agent workload is that scaling isn’t only a compute problem — state and isolation become architectural problems too. I’ve been using Workers for an MCP/Open Data project, and the isolate model works extremely well for short-lived tool execution. I’m curious where you see the boundary between isolates and persistent sandboxed environments once agents start maintaining longer-lived state or executing arbitrary code.
An interesting point. AI infrastructure isn't just about supporting larger models; it's about building an architecture that can reliably handle low latency, scalability, security, and global distribution. The organizations investing in those fundamentals today will be better positioned as agentic AI continues to evolve.