Cross-Cutting Findings

Some insights don't belong to a single focus area — they emerge when you look across bootstrap optimization, cron tuning, agent architecture, and behavioral experiments together. This section captures patterns and principles that span multiple domains.

Emerging Themes

Token budget is a system-level constraint

Bootstrap context, cron prompts, sub-agent spawns, and tool descriptions all compete for the same context window. Optimizing one area in isolation (e.g., compacting AGENTS.md) can be undone by growth in another (e.g., adding cron state). Token budgeting needs to be holistic.

Prompt-level instructions are suggestions, config-level controls are enforced

Models routinely ignore prompt-level constraints under pressure — "don't use this tool" in AGENTS.md gets bypassed when the model thinks using the tool would help. Real enforcement requires config-level controls (tool allowlists, model routing, session isolation).

Measurement before optimization

Token audits, cron timing analysis, and compliance measurement all share the same insight: ad-hoc observations ("AGENTS.md feels too long") don't produce actionable changes. Every optimization area needs a measurement baseline with concrete numbers before changes can be evaluated.

Status

Growing. This section will expand as patterns emerge across the other focus areas. Each section's findings that have cross-cutting implications should be referenced here.