Last year, I was skeptical about Walrus. The mainnet has been running for nearly a year, with token prices fluctuating, and although the Sui ecosystem is lively, I haven't seen any obvious signs of explosive growth in the storage sector. At that time, I even reduced my position by half, thinking this might be another project with a loud noise but little substance.

But less than half a month into 2026, I bought back in. It wasn't driven by market trends but by some recent practical applications and project updates that completely changed my perspective.

What triggered me the most was a technical sharing about verifiable AI data — discussing how the Sui tech stack can be combined with Walrus to achieve end-to-end verifiable and provable AI data. After reading it, I immediately tested a few datasets locally. Previously, I thought that combining AI and blockchain was purely idealistic, after all, data sources and quality are controlled by centralized giants, right? But now, it's different. Walrus allows data blobs to be stored directly on-chain, with transparent sources and tamper-proof. For AI model training, there's no need to worry about data poisoning or unknown sources anymore. This kind of reliability truly changed my view of the entire direction.

Besides data verification, I also tried Walrus Sites — the website building tool based on decentralized storage. The whole process was surprisingly simple: upload code, generate a link, run entirely on-chain, without traditional servers or hosting services. It seems like a small feature, but behind it represents the real implementation of storage needs.

It takes time for a project to go from concept to application. Walrus is currently in this stage, quietly building the infrastructure.