Serial Alice is a tool for generating verifiable evidence of energy consumption for AI workloads. It measures compute energy directly at hardware sources, including GPUs via NVML, CPUs and DRAM via RAPL, and PSUs via Redfish. Inside confidential VMs, it can bind an Intel TDX quote to the samples. Each measurement is cryptographically signed at the point of capture and anchored on the Polygon mainnet blockchain. This process creates a tamper-evident record that can be independently verified by any party without requiring an account or API key. The workflow begins by wrapping an AI training run or inference job. Power samples are collected from the hardware sources, and raw readings are hashed. These readings are cross-checked against independent sources to calculate a trust score. The measurement is then canonicalized, hashed with SHA-256, and signed using hybrid certificates that combine Ed25519 and ML-DSA-65 post-quantum signatures. Action records can bind agent identity, runtime context, and a declared mandate reference to the signed content. Certificates are compiled into a SHA-256 Merkle tree, and the root is anchored on the blockchain. Each certificate becomes a public, QR-scannable passport that can be retained as an offline evidence bundle. The tool is designed to refuse unverified data; a number without a certificate is not used in its analysis. It provides recommendations based on measurements, listing the basis certificates for each claim. This supports compliance frameworks such as the EU AI Act Article 53 and CSRD ESRS E1 reporting, allowing auditors, regulators, and customers to verify energy, carbon, signature, and on-chain anchor independently.