Spending apps forget what you taught them because your rules live in someone else's database or a localStorage blob. And when an agent meets an unnamed hex address on-chain, it either guesses or gives up. I built RecallSpend for the @tatum_io x @WalrusProtocol Total Recall hackathon on @SuiNetwork — a Sui spending tracker with no database: • One-time sync via Tatum Sui GraphQL (6 credits, 1 bulk write). Every page load after that reads strictly from Walrus Memory for 0 credits. • Every spend, human correction, and agent payment is an immutable, append-only line in Walrus Memory. Corrections never overwrite observations, so the app always knows what it thought before you taught it otherwise. • When it meets an unknown merchant, you can correct it for free or let the agent buy the label for 0.01 USDC over MPP on Sui Mainnet. • Payment and memory writes are split in two steps, and proveUsdcTransfer verifies the 0.01 USDC transfer on Sui before the label is written. • Generalisation (0% → 100%) is scored only on held-out repeat payments you never touched. • Ask Memory cites the exact Walrus lines behind every answer, with no LLM in the loop. The live app is pre-loaded with a synced demo wallet and a pre-funded agent wallet so you can test a real 0.01 USDC mainnet lookup in one click. Website : https://t.co/JZCXf1J0K3 Github : https://t.co/fp5WieJFYo Retweeted by @WalrusProtocol from @groziez(Historical earnings: For fiscal 2026FY (period ended 2026-09-03), Micron reported basic EPS of 75.38, diluted EPS of 74.33, net income of USD 84.969 billion, and revenue of USD 133.188 billion. Consensus expectations: For the upcoming 2026Q4 event, consensus expects EPS of 32.5625; no revenue estimate is available.)