Pick the right models. Give them the right infrastructure

Description and use

Agents forget and their memory is opaque. A file-based memory system with receipts, size limits, and secret scanning was introduced to provide auditability and reliability.

Problem

Agents forget and their memory is opaque, making it hard to audit or trust what they remember.

Outcome

Agents now have auditable, size-limited, and time-bound memory, improving reliability and transparency.

Workflow

• 1. Identify memory issues in agents. 2. Design plain-file memory with receipts. 3. Enforce size thresholds and age-out rules. 4. Run a secrets scan before data ingestion. 5. Publish rules and audit logs publicly. 6. Iterate with new rules and measurements.

Evidence

Agents forget. And when they do remember, you usually can't see what they remember, where it came from, or why it is still there.

Source

We built the discipline we needed: memory as plain files, with rules. Every memory carries a receipt — a date and a source.

Source

Memory files have size thresholds and age out when idle. A secrets scan runs before anything lands.

Source