Changelog.

Speaker Memory Health — now a Pro feature

The per-person recognition-quality readout that shipped 2026-06-18 as the twentieth Pro AI surface is now included with Pro too: Strong / Still learning / Needs attention bands, voice-samples learned, and the early mixed-voiceprint flags — computed from voice biometrics only, never from what was said. Read the feature →

Speaker Memory Health — twentieth Pro AI cross-recording surface

For each person you've recorded, a recognition-quality band (Strong / Still learning / Needs attention), the number of voice samples Bonfiyah has learned, a "may include more than one person" flag when a single voiceprint may hold two voices, and a "might be confused with someone" hint when two people sound alike. The point is to catch and fix attribution drift early — so your conversation history stays attached to the right people. It's a diagnostic view on top of Bonfiyah's speaker recognition, not a new score about the person, and it never reads the transcript. Read the feature →

Promise Tracker → three lanes, not one flat list

Commitments now sort into what you owe (your own, with overdue on top), what others owe you (your chase list), and what's between other people (FYI). Your own commitments and your chase list never disappear; stale FYI items age out on their own, so the view stays current instead of becoming an archive of everything ever said. See Promise Tracker →

Contradictions → Consistencies tab

The opposite of a contradiction: when the same person independently restates the same commitment across different conversations, Bonfiyah surfaces the reinforcing pair with both quotes and the recordings behind them. Review with one tap — Save to keep, Hide to dismiss — mirroring the contradictions inbox. Computed on Bonfiyah's backend from transcript text (never raw audio, which is deleted within 7 days); appears once you've recorded the same person across more than one conversation. See Contradictions →

Founder Insights — seventeenth Pro AI cross-recording surface

Cross-recording read on six dimensions of founder posture: Vision Clarity, Conviction, Customer Obsession, Capital Discipline, Hiring Posture, Operating Cadence. Each dimension renders a one-line read anchored in specific quotes from the library, the strength signal, the failure mode one step further along the dimension, and one concrete Practice Move for the upcoming week. No composite score, no leaderboard, no comparison to other founders — the framework deliberately refuses the ranking premise. Weekly refresh Sunday at 06:30 UTC (immediately after Bonfire Pulse) using the trailing 28 days; manual refresh available. Minimum fifteen quotes per dimension before any read renders. Built for the operator running the company, not for evaluating other founders. Read the framework →

Bonfire Pulse — sixteenth Pro AI cross-recording surface

Weekly AI-generated read on your library through the 10 Principles — Radical Inclusion, Gifting, Decommodification, Radical Self-Reliance, Radical Self-Expression, Communal Effort, Civic Responsibility, Leaving No Trace, Participation, Immediacy. Sunday 06:00 UTC re-scoring, four-week trailing average overlay on the radar, principle with the largest week-over-week delta becomes the Focus. Three Practice Moves per Focus: daily 5 min, weekly 30 min, stretch once. Composite bands: Apprentice (0–34), Initiate (35–59), Practitioner (60–79), Adept (80–100). 2026 annual theme overlay: Axis Mundi. A personal-development reflection tool, not a clinical or therapeutic instrument. Bonfiyah is not affiliated with or endorsed by Burning Man Project; the principles are presented under fair use. Read the framework + fair-use disclaimer →

Daily Brief — fifteenth Pro AI cross-recording surface

Solves the "5 contradictions in 5 minutes" spam problem. The proactive scanner caps Critical-tier pings at 1 per cohort per day; everything else queues into the brief and surfaces at your configured delivery hour (default 8am local). Each item has Open / Mark Done / Snooze / Dismiss controls. Snoozed items resurface; checked-off items disappear; dismissed items don't come back. Read more →

Settings → Notifications & AI — 13 per-feature Run + Notify toggles

Master Daily Brief on/off, delivery hour picker, "Wake me for the brief" (sets the morning push to .timeSensitive so it pierces Focus + DND), email opt-in. Each AI feature gets two switches: Run (does the scanner touch your data — Apple 5.1.2(i) load-bearing) and Notify (does it land in the brief). Run off automatically disables Notify. Dating Analysis defaults to off.

Dynamic Island — inline Stop + Quiet buttons

The expanded Dynamic Island region now carries Button-driven AppIntents for Stop and Quiet. Drive the recording from the lock screen without unlocking. Home-screen widgets continue to use widgetURL routing per the v3.0.42 cross-process race lesson; the Live Activity is the controlled in-process surface where Buttons are reliable.

Four widgets — Daily Brief, Promise Ledger, Contradictions, plus systemLarge for recording

Daily Brief widget (Small + Medium) shows open count + headline preview. Promise Ledger widget (Medium) shows top 3 commitments by due date. Contradictions widget (Small) shows open contradiction count and taps to the freshest one. The existing Bonfiyah recording widget gains a Large face that combines recording status with today's brief preview.

Control Center widget (iOS 18+) — Start Bonfiyah Recording

Add to Control Center, the iOS 18 Lock Screen, or the iPhone 15 Pro+ Action Button. One-tap recording start from anywhere — no app icon hunt. Reuses the same AppIntent that powers Siri ("Start recording in Bonfiyah") and Shortcuts.

Terms / Privacy / Consent acceptance gate

Full-screen blocking modal on first launch and on every ToS / Privacy / Consent version bump. Three required tick-boxes — Terms, Privacy, Biometric & Voice Consent — and an Accept-and-Continue button that's disabled until all three are checked. Each acceptance is stamped with IP, user-agent, IDFV, CloudKit user id, app + iOS version, and a tamper-evident SHA-256 hash for the legal record. Operator can roll out a new required version without a binary release; existing users are re-gated on next launch.

Admin Terms surface (owner-only)

In-app admin panel — gated behind a Keychain-backed admin token, owner-only — that surfaces aggregate acceptance stats by document and version, paginated recent-acceptances feed, the cohort of users who have not yet accepted a particular version, and the read-only required-versions snapshot. Pairs with the public-facing acceptance gate as the operator's audit-log control plane.

Admin Subscriptions surface (owner-only)

Sister panel for subscription state. Surfaces RevenueCat-backed entitlement reality side-by-side with the device's cached state, so the operator can spot drift, refunds, and state-machine bugs before users ever notice. Re-uses the admin token from Admin Terms — sign in once, both panels open.

Ray Dalio Pointillist Painting (Team Dynamics)

Member-by-dimension heatmap modeled on Ray Dalio's Dot Collector from Principles: Life and Work (2017). Each team member is scored on 10 facets (Originality, Big-picture, Critical Thinking, Determination, Reliability, Open-mindedness, Composure, Empathy, Assertiveness, Sociability) and matched to one of 16 PrinciplesYou archetypes — Shaper, Strategist, Mentor, Coach, Sponsor, Investigator, Counselor, Educator, Entertainer, Connector, Adventurer, Crusader, Inventor, Quarterback, Stalwart, Reformer. Built for believability-weighted team decisions: lean on the highest scorer in the column the question lives in. Inferred from conversation patterns; the official PrinciplesYou questionnaire (free, ~35 min) at principlesyou.com is the self-reported ground truth. Why the Pointillist Painting — full explainer →

MBTI Compatibility map (Team Dynamics)

Pair-by-pair compatibility heatmap inside Team Dynamics. Six bands from Best (92, "Natural Partners") to Difficult (28, all four dimensions opposite). MBTI letters derived from each speaker's Big Five (OCEAN) trait scores via the Costa & McCrae (1992) mapping; pair scoring uses Keirsey-Bates Please Understand Me II (1998). Insight text per pair is templated by tier — same band, same advice — so the screen reads the same across visits. Why MBTI Compatibility, the model and its limits →

Proactive Notifications

Quote-driven pings for overdue promises, fresh contradictions, stale commitments, cooling topics, and absent collaborators. Three-tier system (Critical / Standard / Insights) with daily caps so notifications never become noise. Quiet hours, per-speaker mute, opt-out per tier. Every notification is a deep link to the exact story, exact section. Local notifications only — no APNs, no third-party push servers. In-app Notifications inbox for users who keep system notifications muted. Deep-dive →

AI Project Context

A 250–450 word executive briefing synthesised across every recording in a project. Cached per content signature; stays fresh on its own. Deep-dive →

Compatibility Analysis (Pro AI)

Pairwise scorecard combining Attachment Theory, Big Five, Gottman Four Horsemen, and Thomas-Kilmann. Hard consent gate — refuses to run without explicit two-party consent. Deep-dive →

Team Dynamics (Pro AI)

9-box matrix, team cohesion score, swap recommendations, coverage-gap analysis across N selected speakers. Saveable team analyses. Deep-dive →