The next wave of storage demand is agentic: every AI agent needs persistent memory, retrieval corpora, and provenance records - sovereign, encrypted, content-addressed. Sia is unusually well-suited to be that layer. The concrete idea for when the program-strategy work resumes: orient SDK examples, docs, and grant themes around agent-memory workloads - durable memory/state patterns for agents, RAG corpora via the S3 interface, and training-data provenance via content addressing. I’ve built a working reference myself - a personal, encrypted, agent-queryable archive on Sia that I use daily - so this comes from running code, not a slide.
The full argument (including thoughts on evolving how development gets funded, with the honest math stated up front) is here: Thinking out loud: Sia's second decade - demand, the agentic economy, and evolving how we fund the future
Posting in Ideas per the category’s purpose, and aware of the summer pause - intended as input to the strategy work, not a proposal needing action today.