Convergence dashboard
Each card below tracks one agentic-coding pattern across the three primary tools. A pattern is converged when all three tools have shipped it. The timeline shows the sequence of adoptions — who first, who followed, who has not yet.
3 patterns currently tracked. The registry grows as new patterns converge or as historical patterns get backfilled; expect the count to drift upward, not the existing entries.
Safety 1 pattern
Plan mode
SafetyA read-only phase where the agent reads, analyzes, and proposes a plan without writing files. Mitigates premature action on unclear intent.
Scale 1 pattern
Subagent delegation
ScaleSpawning a child agent with its own context and tools to handle a bounded sub-task, returning a summary to the parent.
Context 1 pattern
Top-level briefing document
ContextA conventional filename at the repo root (CLAUDE.md / GEMINI.md / AGENTS.md) where authors encode project context for stateless agents.
Interaction no patterns
No tracked patterns in this category yet. This is a coverage gap, not an assertion that nothing belongs here.
Extension no patterns
No tracked patterns in this category yet. This is a coverage gap, not an assertion that nothing belongs here.
Other no patterns
No tracked patterns in this category yet. This is a coverage gap, not an assertion that nothing belongs here.
How to read this
The registry lives at changelog/patterns.yaml.
Each tool's adoption timeline lives at
changelog/tools/<tool>.yaml. The dashboard
joins them at build time — no code edits needed to add new
patterns or new adoption events.
A pattern with convergence_date: null is either a partial convergence (1 or 2 of 3 tools) or an open design space where the right shape is still being argued. Either way, it is information worth surfacing, not hiding.
Coverage is intentionally sparse in the early book — the registry grows organically through quarterly audit cycles. Pattern nominations are welcome via Issues.