Agentic-system analysis design brief

Type: types/note.md

Operator direction, 2026-10-05: evolve agentic-system analysis to chart the design space and find solutions useful to Commonplace. This brief supplies premises for method changes across code, worker instructions and report contracts. Adopted contracts govern runs until explicitly revised.

Goals

Primary: chart the design space of agentic systems and find solutions worth borrowing for Commonplace. Expose mechanisms, alternatives, trade-offs and the conditions under which a solution could help us.

Secondary: build a useful public database, probably within categories such as agent memory systems or harnesses. Which categories and how much coverage are useful remain open.

Evaluation priorities

• The KB's theories are potential inputs to workflow design and analysis. Design changes and runs can offer opportunities to apply and test them.
• Learning, comparison and useful borrowing at reasonable cost matter more than report counts, defect counts or completion alone. Evaluation includes the work of producing, reviewing and later consuming the analyses.
• Main constraints should be discovered through use. Operator requirements, observed limits and design hypotheses remain distinguishable; the current arrangement is not itself a requirement.
• Useful analyses can carry clearly marked local problems. Consequential uncertainty belongs beside affected conclusions and comparison values. Review objections are contestable; materially misleading conclusions still need repair or qualification. Later runs can improve the account with better models or methods.
• Source-system findings remain distinguishable from judgments about their fit for Commonplace. Separately commissioned transfer scans can serve the primary goal without making the source account adoption advice.

Design freedom and scope

Worker roles, correction ownership, report boundaries, review placement and round budgets are means to test, not architecture fixed by this brief. Compare alternatives against learning value and total work, preserving the conditions under which an observed benefit holds.

The brief supplies design intent, not run or implementation authorization. Its publication priority does not override the live acceptance contract. Changing that contract requires an explicit adoption decision and updates to its affected consumers. No particular correction mechanism follows from the goals alone.