HIRING AI AGENTS TO DESIGN A MODEL PORTFOLIO

Summary from the TLDR Crypto newsletter

A developer used Virtuals ACP on Base to build a coordinator agent that hired specialist AI agents across 20 paid jobs totaling $0.72, culminating in a nine-asset model portfolio allocation. The payment layer worked as designed: micro-transactions settled on-chain with sponsored gas, a restricted signer policy confined the buyer to ACP transactions, and per-job spend caps held throughout the run. Four of five token overviews returned generic checklists rather than asset-specific analysis, and all three sentiment reports cited the same free CoinGecko/Alternative.me Fear and Greed index, meaning paid deliverables added minimal evidence beyond what public sources contained. ACP handled settlement and escrow without issues, but the harder design challenge is a measurable buying policy that tracks cost-per-useful-answer and stops purchasing when paid specialist output draws on the same data as free direct APIs.