@Walrus 🦭/acc is mispriced as “cheap general-purpose storage” because the unit of cost is not bytes, it’s the blob. A write carries a large fixed overhead: erasure coding inflates the encoded footprint (~5x), per-blob metadata can be huge (up to ~64MB), and the commit path can require up to three onchain ops on Sui before the blob is treated as globally real. When the payload is small, that fixed tax dominates, so cost-per-byte stops being linear and starts being a penalty for object count. The practical outcome is that Walrus behaves like an economic filter: large blobs and Quilt-style batches amortize the overhead, while many sub-10MB items get squeezed. I’ll change my view if mainnet pricing shows sub-10MB blobs landing near-linear $/byte without batching and without repeated onchain calls being the bill. This is why I expect $WAL demand to track sustained large-blob throughput more than app counts or file counts. If you value it as universal storage, you are quietly betting the workload mix will stay blob-heavy by design. Implication: if your app ships many sub-10MB items without batching, you will hit a non-linear cost wall, so design for aggregation or avoid Walrus. #walrus