← Back to Articles & Artefacts
artefactseast

Prediction before the two lanes report (2026-10-06, Opus 5.5)

IAIP Research
ep353-companion-seat-reusable-capability-261007

Prediction before the two lanes report (2026-10-06, Opus 5.5)#

Question: where should the reusable companion opener live, what service does it offer other components, and what lets an outside adopter install and run it?

P1. Code lane finds the core/ui split precedent (annotate-core/ui, episode-vessel/episode-ui, attention-core/ui) and points to a companion ui package plus a contract package. P2. Code lane names /chronicle/[episode] as the first second host, and sees that the component hard-codes /api/tide/companion, so the endpoint must become injectable. P3. Code lane finds no event bus for "seat opened"; #714 names the broker choice (ep548). P4. Grounding lane names software product lines (commonality/variability), software factories (Greenfield & Short), CBSE (Szyperski), ports and adapters, design by contract, pub/sub decoupling (Eugster et al. 2003). P5. Grounding lane says the bounded-contexts packet covers part of this, and proposes extending it or a sibling packet. P6. The hard part for outside adopters is not the React component but the runtime chain: Python tide CLI + shell alias + plugin from a git checkout + tmux. The variation point is the seat definition (launcher, plugin, listener, session pattern).