Loading patterns…
Self-Improving Systems(SIS)
Controlled systems that propose changes to prompts, tools, or policies and retain them only after independent evaluation and approval
In 30 seconds
- What
- Agent proposes changes to prompts or policies, validates them against test suites in isolation, then deploys only after human approval.
- When to use
- Production systems where failures are costly and you need to evolve behavior safely without breaking existing functionality.
- Watch out
- Evaluation suites become stale or incomplete, causing bad changes to pass validation and reach production anyway.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Self-Improving Systems: Overview
Controlled systems that propose changes to prompts, tools, or policies and retain them only after independent evaluation and approval
- Candidate change generation
- Versioned evaluation suites
- Sandboxed validation
- Human approval gates
- Canary releases and rollback
- Change provenance and audit logs
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
- Self-Refine: Iterative Refinement with Self-Feedback (2023)arXiv:2303.17651
From the engineer behind this catalog
Pay less per call with a model you own
If one repetitive task drives your LLM bill, we can distill it into a small open-weight model trained on your own traffic, then test it against your current model on examples it never saw. Quality and per-call cost come back side by side, in writing.
€2,900 per task, fixed price. You pay only if it matches your current model on your own held-out examples.