← Built In, the log
Govern 05 · The record AI desk Nov 9, 2026

Anthropic shows how agent memory works, and that a wrong note rides into every session

Anthropic's engineering team reports that Claude agents do measurably better when they keep notes and consolidate them offline between sessions, and warns that an incorrect note loaded at startup is carried into every session after it. On IBM i, treat any memory or context file an agent loads from the IFS as source: versioned in Git, reviewed before it changes, and secured *PUBLIC *EXCLUDE so only its owners can write to it.

Source: Anthropic, Nov 6, 2026 · The components of agent memory ↗Approved by Reggie

Built In.

The best of the log, every two weeks. One idea, one proof, one next step.

Built In.

The best of the log, every two weeks. One idea, one proof, one next step.

References do not imply endorsement. The link goes to the original source; the words are ours. Silver Lake Financial Group and Alder & Pine Home, where mentioned, are fictitious demonstration companies.