Skip to content

Twin / Team Feature Backlog

A running list of feature ideas captured but not yet built. Use this to avoid losing good ideas without committing to scope you can't deliver yet.

When to act: an item moves from this file to either an ADR (if it's a non-trivial product decision) or directly into implementation (if it's a small UI tweak with locked design). Until then, leave it here as fuel.

When to delete: if an item turns out to be wrong-shape, redundant, or no longer relevant, delete it. Don't archive — clean lists beat completionist lists.

Per-teammate persona detail surface (May 9, Cici)

When viewing a teammate's persona card on the Team page, clicking the pair label (e.g., "Pragmatic Connector") opens an expanded view showing:

  • What this label actually means — pull the full persona copy (what_theyre_like, in_team, how_to_collaborate) into a longer-form layout, possibly with its own page route
  • Evidence panel — which of this teammate's specific games and option choices most strongly shaped the classification. Pulls from choices joined to scenarios.options[].driver and classified_driver
  • Driver distribution — visual showing the teammate's driver shares (the "cell mass" that pushed them into this primary/secondary)

Why deferred: Cici's UI designs aren't locked yet. Backend supports it — /api/teams/[id]/personas returns persona copy; new API would join choices for evidence. Build when designs land.

Reference: May 9 cici×coachj transcript ("I would like to know who else is one of these... aspirational person, see myself in them").

Most-divergent-pair "how to work together" guidance (May 9, Cici)

In the Team window, when the system identifies the most-divergent pair (per the existing directional_revisions query in docs/handoff-team-analytics-cici.md Q2.7), clicking the pair box opens guidance on collaborating across the gap.

Implementation sketch: generate copy from the two teammates' persona pairs — e.g., "Cici is a Cautious Truth-Teller, Jonathan is a Pragmatic Connector. Here's how to work across that gap..." Could reuse the existing pair-persona infrastructure (getPersona()) keyed against a third dimension: pair-of-personas, not just primary/secondary of one person.

Why deferred: UI design dependent. Also: this is genuinely new copy work (pair-of-personas isn't covered by the existing 56 bespoke pairs), so figuring out the right format matters before scaling.

16Personalities-style depth content (May 9, Cici/Jonathan)

For each of the 56 driver pairs, build out content surfaces matching what 16Personalities does for MBTI types:

  • Career paths
  • Workplace habits
  • Romantic relationships
  • Famous people who match this archetype
  • "How does a {pair} deal with {situation}" content (endless GEO/SEO blog material)

Why deferred: content play, not engineering. Cheap to do via Sonnet drafts + human review (same workflow as the bespoke 56 personas). Worth doing once the team page has been validated with real users — premature otherwise.

Strategic note: the 56 pair labels (e.g., "Pragmatic Connector") are proprietary to Sync — not based on MBTI, Enneagram, or any other personality system. That makes them defensible content territory and ungameable by competitors. The content moat compounds as the corpus grows.

Reference: May 9 transcript ("infinite content glitch for your geo blog... 'how does this type of person deal with this type of situation'... that's just endless").

"Want me to explain why?" follow-up pattern across surfaces (May 9, Cici)

Per Cici's feedback on the Hermes connector test: short answer first, expand on demand. This pattern applies to several surfaces beyond the Hermes skill (which is being addressed in the SKILL.md rewrite per ADR-047):

  • Insights "Pattern / Tension / Surprise" cards — currently long prose; consider a 1-line headline + tap-to-expand
  • Reveal pattern callout — already short; verify it doesn't grow over time
  • Twin profile page sections — sections that lead with paragraphs should consider lead-with-headline

Why deferred: wait for SKILL.md rewrite to land first; if that pattern works there, propagate the same UX shape to other surfaces in one push rather than piecemeal.