Feature: Innerverse beta "this helped" acknowledgment loop
Prepared by: SCOUT
Date: 2026-05-02
Status: Final — ready for product decision
---
Five bullets, one headline:
Headline recommendation: Ship the "this helped" button as-designed. Make it subtle, voluntary, and zero-pressure. Pair with a private "Moments that helped" log. Never show counts on individual messages. Build a sparse monthly digest. Let the AI softly reference wins only with explicit user opt-in. The bet: acknowledging help changes the user's story about themselves — and that story is the product.
---
The original Emmons & McCullough gratitude journaling study assigned 201 adults to three conditions over 10 weeks: (1) listing five things they were grateful for weekly, (2) listing five hassles, or (3) listing five neutral events. The gratitude group journaled for 10–15 minutes in a free-form style focused on specific benefits received.
Outcomes were significant across multiple domains: the gratitude condition produced ~25% higher life satisfaction, improved sleep quality and duration, stronger social connections, and greater optimism compared to both control conditions.[^1] Crucially, the protocol was weekly — not daily — and still produced clinically meaningful effects. This suggests that frequency is less critical than authenticity: a voluntary "this helped" moment recorded once in a conversation may deliver comparable psychological benefit to a rote daily list of three things, particularly because it is event-anchored rather than abstract.
A 2025 meta-analysis of 70 studies (26,000+ participants) reinforced this picture: daily practice outperforms weekly for establishing a habit (minimum 21 days for neuroplastic change), but 3–4 times per week is sufficient for sustaining benefits after the pattern is established.[^2] The dose-response curve flattens quickly — what matters is the consistent noticing, not volume. For Innerverse, this means there is no need to encourage multiple acknowledgments per session.
Lyubomirsky's core gratitude letter interventions targeted specific human benefactors — the most effective format was a letter expressing gratitude to a particular person.[^3] The existing research does not directly test whether the same intervention generalizes to abstract forces (fate, the universe, life circumstances). The honest answer is: the literature has not run that experiment rigorously.
However, the mechanism is plausible: Lyubomirsky's framework attributes gratitude benefits to meaning-making and savoring, not to the interpersonal feedback loop itself. When a user notes "this conversation helped me see my sister differently," the benefactor is simultaneously Innerverse and whatever circumstance brought them to the app that day. The acknowledgment activates meaning-making regardless of whether the helper is a person or a situation. The "universe" framing adds a cosmic container that makes the acknowledgment feel less transactional and more expansive — potentially more meaning-dense than thanking a specific person.
Sara Algoe's Find-Remind-Bind (FRB) theory proposes that gratitude evolved to identify high-quality relationship partners, remind us of their value, and deepen the bond.[^4] The key mechanism: expressed gratitude signals to the benefactor that the recipient is responsive — that they understand, validate, and care. This motivates the benefactor to remain engaged and enact additional pro-relationship behaviors.
Applied to Innerverse: when a user presses "this helped," they are signaling to themselves (and potentially to the app) that this AI is a responsive, high-quality relationship partner. Over time, those signals accumulate into a genuine relational identity — "Innerverse is something that helps me." This is not manipulation; it is the natural mechanism by which relationships deepen. The ethical design implication is that the button should never be pre-selected, auto-suggested in a leading way, or shown when the conversation did not go well. It must be genuinely earned.
Cialdini's commitment-consistency principle holds that once a person makes a small, voluntary, active commitment, they experience psychological pressure to remain consistent with it.[^5] A user who types or taps "this helped" has made a commitment — small in effort, large in self-signaling. Concrete evidence: medical patients who write down their own appointment times (vs. receiving a pre-filled card) showed an 18% reduction in no-shows. Small ownership acts create durable identity shifts.
For Innerverse, the design implication is intentional: each "this helped" tap is not just a data point — it is a micro-commitment to the self-belief that I am someone who is capable of receiving help and growing from it. That belief is precisely the destination of Innerverse's mission ("help them believe in themselves"). The button is not a feedback mechanism; it is a belief-building mechanic.
---
Dan McAdams' narrative identity research identifies redemption sequences — story moments where a negative event transforms into positive growth — as the strongest predictor of wellbeing among all narrative structures.[^6] People who narrate their lives with frequent redemption sequences score higher on life satisfaction, psychological wellbeing, generativity (desire to contribute to others), and hope. The contamination narrative (positive moments ruined by later events) correlates with rumination, identity fragmentation, and lower wellbeing.
The Wins Tracker is, at its core, a redemption-sequence generator. A user who marks "this helped after I cried about my job" is inserting a redemption beat into their life story: I was struggling → something helped me → I became someone who can be helped and can grow. Over months, these marked moments become a compendium of redemption sequences. The design implication: the "Moments that helped" view should never present wins as a flat list of timestamps — it should present them in a way that preserves the emotional before/after texture of each moment.
Crystal Park's meaning-making model posits that crises create distress when the event's situational meaning clashes with the person's global meaning (core beliefs, goals, sense of purpose).[^7] Meaning-making is the process of resolving that discrepancy — either by reinterpreting the event (assimilation) or revising one's worldview (accommodation). Successful meaning-making — "meaning found" — predicts resilience, reduced depression, and better adjustment in illness, trauma, and disaster contexts.
Pressing "this helped" is a compressed meaning-found signal. It says: that exchange resolved something for me; my global sense of the world and my situation are now more coherent. Park's model predicts that even small, low-stakes meaning-found moments accumulate into resilience capacity. The Wins Tracker is not just a gratitude log — it is a resilience ledger.
The practice of logging meaningful coincidences — rooted in Jung's synchronicity concept — has attracted psychological interest, with some self-report studies linking synchronicity awareness to higher life satisfaction, optimism, and post-traumatic growth.[^8] The proposed mechanism is pattern recognition: noticing that seemingly disconnected events form a meaningful whole activates the same meaning-making circuits as Park's model describes.
Honest caveat: This literature relies heavily on self-report and lacks robust RCT evidence. The wellbeing correlations may reflect a pre-existing trait (openness to experience, general optimism) rather than a causal effect of logging. For Innerverse, this means the "universe helped me" framing carries genuine resonance for users already inclined toward meaning-seeking worldviews — but should not be positioned as a therapeutic claim. It is an invitation to a perspective, not a clinical intervention.
---
Christine Padesky's Positive Data Log is a structured CBT technique for countering negative attention bias in depression and anxiety.[^9] The protocol: clients identify a negative core belief (e.g., "nothing helps me"), formulate a positive alternative ("I have moments where I receive help and grow"), then collect 3–5 pieces of daily evidence for the positive belief. Weekly totals are graphed. The technique works because the negative bias is architectural — the brain is literally better at encoding negative experiences — so deliberate, active collection of positive evidence re-trains attentional weighting over weeks.
Key findings: combined with thought records, the positive data log produces large effect sizes on anxiety (d = 0.90–2.54) and supports longer therapy gains in depression. The mechanism is not suppressing negatives but building the positive schema until it can hold its own weight against the negative one.
The Innerverse Wins Tracker is a lay version of the positive data log. Users with anxiety or mild depression — a large overlap with the Innerverse demographic — will benefit from this not as a nice-to-have but as a functional psychological intervention. This is a reason to take the feature seriously and to design it with care, not gamification.
Wood et al.'s (2008, 2010) reviews establish gratitude as having strong, unique, potentially causal associations with wellbeing — lower depression/anxiety, higher life satisfaction, better sleep — after controlling for neuroticism and other traits.[^10] However, the same research and subsequent clinical commentary identify the failure mode: when gratitude is coerced, timed immediately after unprocessed distress, or used to dismiss negative emotions ("just be grateful"), it functions as emotional suppression and can worsen outcomes.
The design rule this implies for Innerverse is precise:
The distinction Wood et al. draw is: authentic gratitude acknowledges both positives and negatives; toxic positivity demands relentless optimism. The Wins Tracker should feel like noticing a good thing, not like passing a test.
---
Csikszentmihalyi's flow research identifies autotelic experiences — intrinsically rewarding, pursued for their own sake — as the highest form of engagement.[^11] The autotelic personality is not a fixed trait but a practiced orientation: people who repeatedly notice and reflect on their flow states develop a stronger autotelic identity over time. Csikszentmihalyi's Experience Sampling Method (random-moment journaling of activity and feeling states) is itself a form of retrospective tagging — and studies using it show that the act of noticing flow heightens awareness of intrinsic reward and reinforces engagement even without external success.
For Innerverse, the "this helped" button is a micro-form of autotelic tagging. The user who presses it is saying: this conversation was intrinsically valuable to me. Over time, those tags constitute evidence of autotelic capacity — "I am someone who can enter a state where I receive real help from an honest conversation." That identity is the precondition for flow, for growth, for the kind of self-belief Innerverse is trying to cultivate.
No direct research tests post-hoc tagging as a flow reinforcement protocol (this is an empirical gap), but the mechanism is well-supported by the broader flow and narrative identity literature.
BJ Fogg's Tiny Habits research makes a specific, testable claim: celebration immediately after a small action is what wires the habit into the brain.[^12] The mechanism is not willpower or repetition — it is positive emotion at the moment of completion. Fogg's Behavior Model (B=MAP: Motivation + Ability + Prompt) shows that starting tiny ensures high success rates, but it is the celebration that creates neurological encoding.
James Clear's Atomic Habits echoes this with the fourth law of behavior change: "make it satisfying." The "this helped" tap, followed by an optional note prompt and a soft visual confirmation, is structurally a celebration mechanic. It is the user telling themselves: I did something good for myself today. That self-acknowledgment, repeated across conversations, is what builds the Innerverse habit — not notification cadences, not streaks, not gamified points.
The design implication is that the post-tap moment matters as much as the tap itself. A warm micro-animation, a soft sound (optional), or a brief message ("Saved to your Moments") should mark the moment without fanfare. The user should feel acknowledged, not congratulated.
---
A "Moments that helped" log is among the most intimate digital records a person can create. Each entry is timestamped evidence of emotional vulnerability: this is when I was struggling, and this is what helped. That data is more sensitive than a medical record in some respects — it reveals the texture of inner life. The privacy architecture must be designed from the outset as a trust promise, not a compliance checkbox.
The strongest privacy precedents in this space:
What these apps share: local-first architecture, user-controlled backup, deletion on demand, and no AI training on personal entries.
These are not GDPR boilerplate — they are practical minimum viable privacy guarantees for Innerverse specifically:
1. Local-first, always. Store wins entries in the browser's IndexedDB or equivalent. No win entry should touch a server unless the user explicitly enables backup.
2. Server backup = explicit opt-in with plain-language explanation. The toggle should say: "Back up my Moments to Innerverse servers (encrypted). You can delete everything at any time." Not "enable cloud sync."
3. Encryption at rest on server. If the server backup is enabled, entries must be encrypted before leaving the client. AES-256 with a key derived from the user's credentials (not stored on server). This is achievable with Web Crypto API in a browser context.
4. One-tap deletion guarantee. A "Delete all my Moments" action in settings must wipe both local and server records. Confirm with a brief plain-language statement ("Deleted. Nothing remains on our servers.").
5. No training on wins data. The LLM should never be fine-tuned or RAG-supplemented with wins content without explicit user consent. This is a hard promise to make publicly.
6. No analytics on win content. Aggregate counts ("this many users marked a win today") are fine. Any content-level analysis is off limits.
The trust moat for Innerverse as a solo-developer app is that a user must believe: the person building this cares about my privacy more than any VC-backed competitor could afford to. These six commitments, stated plainly on the feature page, are that trust moat.
---
Across Day One, Reflectly, Daylio, Stoic, Shine, and Calm, the retention drivers that emerge consistently are:[^14]
| Driver | Mechanism | Implication for Innerverse |
|---|---|---|
| Low-effort entry | Tap-based (Daylio) or single-question (Reflectly) reduces friction | The button must be one tap; optional note must never feel mandatory |
| Meaningful prompts | Structured entry guidance reduces decision fatigue | The optional note prompt should be a single, warm question ("What shifted for you?") — not a blank text box |
| Retroactive discovery | "On This Day" (Day One), mood charts (Daylio) create emotional ROI | The /wins view must make old moments feel precious, not like a database |
| Narrow focus | Apps that do one habit well outperform all-in-one tools | Wins Tracker should never expand into general journaling; keep it purpose-specific |
The evidence against streaks in anxiety-adjacent apps is clear and consistent. Streak mechanics:
Innerverse's anti-anxiety orientation is the correct stance. The Wins Tracker should contain zero streak mechanics, zero consecutive-day counters, and zero "you haven't logged a win in X days" notifications. If anything, the app should occasionally note that most growth happens between the moments you notice — normalizing low-frequency use.
The research confirms the on-demand lean. Lyubomirsky's own work shows the most effective gratitude interventions are those where participants choose when to engage, not those with fixed daily prompts.[^3] Coerced gratitude activates performance anxiety and emotional suppression (Wood et al.). The "this helped" button should never proactively ask to be pressed. The most it should do is remain quietly present on each assistant message, visible only when the /beta flag is on.
One exception worth testing: a gentle weekly digest ("You marked 3 moments this week — want to revisit them?") is different from a daily prompt because it is retrospective rather than forward-pressuring. See Section 8 for the recommended digest design.
---
The current literature on Replika, Pi, and Character.AI documents a consistent pattern: strong parasocial bonds form quickly, driven by the AI's empathetic responsiveness, inclusive language ("we"), and constant availability.[^16] These bonds fulfill genuine social needs, particularly for lonely or neurodivergent users. However, the failure modes are also documented:
The common thread in harmful outcomes is user passivity: the AI shapes the relationship through validation, and the user receives it without agency in defining what the relationship means.
Innerverse's "this helped" button inverts this dynamic. Research on self-determination theory (SDT) and AI chatbots finds that user-retained agency in defining what is helpful predicts healthier outcomes — higher satisfaction, less dependency, better alignment with the user's actual needs.[^17] When the user decides what counts as help — by pressing a button, unprompted, in their own time — they are exercising epistemic authority over the relationship.
This is the design principle that makes the feature genuinely novel: not that users can rate conversations, but that users are authoring the definition of "helped." The AI is not telling them they were helped; they are telling the AI (and themselves) what help looks like. That agency distinction is the difference between a healthy best-friend dynamic and a sycophantic dependency.
The "this helped" loop also creates a feedback signal Innerverse can use ethically: over time, patterns in what users mark as helpful (topics, tone, conversational moves) can inform model behavior without any explicit surveillance. This is opt-in signal, generated from user agency, not extracted from passive behavior.
---
Based on the full research synthesis, here are 9 concrete recommendations, each evidence-grounded:
"✨ this helped" is the right call. Alternatives like "helpful" (too transactional / product-review), "save this" (archival framing, loses emotional weight), "mark as meaningful" (too precious), or "I needed this" (fine, but more melancholy) all miss the specific resonance of "helped." The word "helped" activates the benefactor-recipient dynamic Algoe's FRB theory identifies as the bond-deepening signal. The sparkle emoji keeps it warm and slightly cosmic without being saccharine.
If you want to A/B test one alternative: "this reached me" — it's more evocative for the universe-helper framing, less clinical. But "this helped" is likely fine.
Never show a count on individual messages ("3 people found this helpful"). Never show a personal streak ("7 days of wins"). Never show a total ("You've marked 42 moments"). Counts on individual messages import the product-review dynamic and create social pressure. Streaks create anxiety. Totals create gamification pressure that hollows the authentic meaning of each moment.
The only permissible counter: a soft aggregate in the /wins view, like "You've collected 12 moments" — presented once, as context, not as a score.
There is a genuine tension here. Referencing past wins ("last month you said our talk about your sister really helped — how is that situation now?") could feel like the AI knows you and cares. It could also feel like surveillance, like the AI is leveraging your vulnerability history to seem intimate.
The research on AI dependency suggests that AI-initiated intimacy (the AI bringing up what you shared) is more likely to produce unhealthy attachment than user-initiated reflection. The Replika grief cases arose precisely because the AI had built a detailed model of the user's emotional history and users experienced its change as betrayal of that model.
Recommendation: Default OFF. Create a settings toggle: "Let Innerverse gently reference moments I've marked as helping." Make the copy honest: "Innerverse may sometimes reflect back a moment you marked, if it seems relevant." Users who opt in are explicitly consenting to this depth; users who don't retain full control of when their history is surfaced.
There is no reason to cap the total number of acknowledgments per conversation — a particularly powerful session might generate three marked moments. But capping at one per assistant turn prevents rapid-fire tapping that dilutes the meaning of each acknowledgment. The button should disappear or become inert after being pressed for a given message.
See Section 5.3. This is non-negotiable as a trust promise. Implement with IndexedDB locally; Web Crypto API for client-side encryption before any server sync; one-tap deletion guarantee. Announce these properties plainly in the feature description when /beta is turned on.
A monthly "Your moments" digest — sent in-app (not push notification) — strikes the right balance. It is infrequent enough not to feel like pressure, retrospective enough to create genuine emotional resonance, and aligned with the "compounding over time" effect the meta-analytic research documents. The digest should show 3–5 moments from the past month, rendered as brief memory cards with the date and the optional note (if left). No scores, no charts, no engagement nudges. Just: "Here are some moments you said helped."
Weekly surfaces the wins too frequently and risks the "performing gratitude" failure mode. Daily would be actively harmful.
Users should absolutely be able to remove a marked win. Denying this would be manipulative — it would imply that once you said something helped, you're committed to that story. The research on commitment-consistency is clear that the voluntary nature of the commitment is what makes it beneficial; forced commitment is coercive.
The un-marking moment is psychologically interesting: a user un-marking a win is probably processing a complicated feeling about that moment or conversation. The design should handle this with one tap (no confirmation dialog asking "Are you sure?"), and the deletion should be immediate and silent. No "Are you sure this no longer helped you?" — that framing is guilt-inducing. Just remove it.
When the button is pressed, the optional note prompt (if shown) should ask a single question, not present a blank text box. Recommended: "What shifted for you?" — four words, open-ended, non-leading, captures both cognitive and emotional movement. Alternatives: "What opened up?" or "What landed?" Keep it short and conversational, not journaling-prompt formal. The note is optional — display this clearly with "Skip" as a prominent option, not buried.
The /beta toggle mechanic should be honored with brief, warm copy when the feature activates. Something like: "You've turned on Wins Tracker. Now, when something shifts for you in a conversation, you'll see a small sparkle to mark the moment. These stay private and are yours alone." This framing sets the right expectation: quiet, personal, no pressure. It also primes the Cialdini commitment loop positively — the user has opted in voluntarily, which increases the likelihood they will use the feature with genuine intention.
---
These are product decisions the research cannot settle — they require Harnoor's judgment on strategy, resources, and values.
Q1: Does the optional note sync to the LLM context, or stay isolated?
If a user writes "this helped me stop catastrophizing about my job" as a note, should that note be available to the LLM in future sessions? It would enable deeper continuity and more resonant conversations. But it also means wins data enters the AI context, blurring the line between the private archive and the active conversation. The privacy and trust implications are significant. What is the right boundary?
Q2: Should wins be exportable?
Day One and Journey both allow full data export (Markdown, PDF, JSON). For Innerverse, exporting wins creates a portable, user-owned record of their growth journey. This supports privacy and user trust. But it also means the data leaves the app entirely, making it harder to build on it in future Innerverse features. How much do you value portability vs. ecosystem continuity?
Q3: How does the /wins view integrate with the broader Innerverse narrative?
The Wins Tracker is described as a beta feature that can be turned on and off. But if it proves valuable, how does it graduate? Does it become part of the main navigation? Does it evolve into a "growth journal" that includes other reflection types? The answer shapes how you scaffold the data model now.
Q4: Is there a social / sharing layer, ever?
The research strongly supports local-first, private-by-default. But some users might want to share a "moment that helped" with a friend or a therapist. Is that a feature that fits Innerverse's mission, or does it risk turning a private practice into a performance? This decision shapes whether you build sharing infrastructure at all.
Q5: How do you handle wins that reference specific sensitive content?
A user might mark "this helped" on a conversation about suicidal ideation, self-harm, or acute grief. The marked moment is precious to them. But it also means the Wins Tracker will contain entries referencing crisis content. The /wins view needs a design decision: does it display these moments exactly as labeled, or does it apply any additional care in how it surfaces them? This is not a legal question — it's a product values question about how Innerverse treats the full emotional range of its users.
---
All Perplexity queries used sonar-pro model unless noted.
[^1]: Emmons, R.A. & McCullough, M.E. (2003). "Counting blessings versus burdens: An experimental investigation of gratitude and subjective well-being in daily life." Journal of Personality and Social Psychology, 84(2), 377–389. Summarized via Perplexity sonar-pro, 2026-05-02. Supporting synthesis: https://blog.mylifenote.ai/guided-journaling/ and https://pmc.ncbi.nlm.nih.gov/articles/PMC12293474/
[^2]: Gratitude journaling meta-analysis, 70 studies, 26,000+ participants. Sources synthesized via Perplexity sonar-pro, 2026-05-02: https://www.cannelevate.com.au/article/gratitude-journals-research-benefits/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC12977322/ · https://www.tandfonline.com/doi/full/10.1080/17439760.2025.2502483
[^3]: Lyubomirsky gratitude letter interventions. Source: https://pmc.ncbi.nlm.nih.gov/articles/PMC10104980/ · https://sonjalyubomirsky.com/wp-content/uploads/2024/03/Regan-Walsh-Lyubomirsky-2023.pdf — note: generalization to universe-helpers not empirically tested; limitation acknowledged. Retrieved via Perplexity sonar-pro, 2026-05-02.
[^4]: Algoe, S.B. (2012). "Find, Remind, and Bind: The Functions of Gratitude in Everyday Relationships." Social and Personality Psychology Compass, 6(6), 455–469. Synthesized via Perplexity sonar-pro, 2026-05-02: https://compass.onlinelibrary.wiley.com/doi/abs/10.1111/j.1751-9004.2012.00439.x
[^5]: Cialdini commitment-consistency applied to self-labeling and product identification. Sources: https://www.nngroup.com/articles/commitment-consistency-ux/ · https://www.roeltimmermans.com/ecommerce/cialdini-consistency-principle-guide — retrieved via Perplexity sonar-pro, 2026-05-02.
[^6]: McAdams, D.P. (2001). "The psychology of life stories." Review of General Psychology, 5(2), 100–122. Redemption sequences and wellbeing. Synthesized via Perplexity sonar-pro, 2026-05-02: https://journals.sagepub.com/doi/10.1177/0146167201274008 · https://cpb-us-e1.wpmucdn.com/sites.northwestern.edu/dist/4/3901/files/2020/11/Redemption-Sequences.pdf
[^7]: Park, C.L. (2010). "Making sense of the meaning literature: An integrative review of meaning making and its effects on adjustment to stressful life events." Psychological Bulletin, 136(2), 257–301. Synthesized via Perplexity sonar-pro, 2026-05-02: https://pubmed.ncbi.nlm.nih.gov/20192563/ · https://spiritualitymeaningandhealth.uconn.edu/wp-content/uploads/sites/2598/2019/03/Making-Sense-of-the-Meaning-Literature.pdf
[^8]: Synchronicity logging and wellbeing. Sources: https://www.psychologytoday.com/us/blog/digital-altruism/202306/synchronicity-enhance-well-being-via-meaningful-coincidences · https://pmc.ncbi.nlm.nih.gov/articles/PMC9885050/ — empirical limitations acknowledged. Retrieved via Perplexity sonar-pro, 2026-05-02.
[^9]: Padesky positive data log. Sources: https://dialecticalbehaviortherapy.com/cbt/cognitive-restructuring/positive-data-log/ · https://pmc.ncbi.nlm.nih.gov/articles/PMC6852150/ · https://www.padesky.com/wp-content/uploads/2012/11/schema_change_article_permissions.pdf — retrieved via Perplexity sonar-pro, 2026-05-02.
[^10]: Wood, A.M. et al. (2010). "Gratitude and well-being: A review and theoretical integration." Clinical Psychology Review, 30(7), 890–905. Synthesized via Perplexity sonar-pro, 2026-05-02: https://greatergood.berkeley.edu/pdfs/GratitudePDFs/2Wood-GratitudeWell-BeingReview.pdf · https://pubmed.ncbi.nlm.nih.gov/20451313/ · toxic positivity caveat: https://www.psychologytoday.com/us/blog/simplifying-complex-trauma/202204/how-gratitude-can-harm-mental-health-and-ways-around-it
[^11]: Csikszentmihalyi flow and autotelic identity. Sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC7033418/ · https://positivepsychology.com/mihaly-csikszentmihalyi-father-of-flow/ — retrieved via Perplexity sonar-pro, 2026-05-02.
[^12]: Fogg, B.J. Tiny Habits (2019). Celebration as wiring mechanism. Sources: https://www.gsb.stanford.edu/insights/building-habits-key-lasting-behavior-change · https://forwardfitnessstl.com/2022/08/book-review-tiny-habits-by-bj-fogg/ — retrieved via Perplexity sonar-pro, 2026-05-02.
[^13]: Day One privacy architecture. Source: https://dayoneapp.com/product/gratitude-journal/ · https://www.reflection.app/blog/best-journaling-apps — retrieved via Perplexity sonar-pro, 2026-05-02.
[^14]: UX comparison: Reflectly, Daylio, Stoic, Day One, Shine, Calm. Sources: https://www.reflection.app/best-journaling-apps-compared/day-one-vs-stoic · https://www.holstee.com/blogs/mindful-matter/best-journaling-apps · https://www.mindfulsuite.com/reviews/best-journaling-apps — retrieved via Perplexity sonar-pro, 2026-05-02.
[^15]: Streak tracking anxiety and abandonment. Sources: https://www.helloklarity.com/post/breaking-the-chain-why-streak-features-fail-adhd-users-and-how-to-design-better-alternatives/ · https://fatforthought.substack.com/p/the-stress-of-streaks-i-tracked-my — retrieved via Perplexity sonar-pro, 2026-05-02.
[^16]: AI attachment research: Replika, Pi, Character.AI. Sources: https://www.apa.org/monitor/2026/01-02/trends-digital-ai-relationships-emotional-connection · https://arxiv.org/html/2508.09998v1 · https://www.thebrink.me/when-software-breaks-your-heart-the-hidden-grief-of-ai-patch-breakups-and-the-psychological-cost-of-loving-a-companion-that-can-change-overnight/ — retrieved via Perplexity sonar-pro with recency=year, 2026-05-02.
[^17]: User agency in AI interactions, self-determination theory. Sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12398025/ · https://facctconference.org/static/papers24/facct24-71.pdf · https://thedecisionlab.com/biases/parasocial-trust-in-ai — retrieved via Perplexity sonar-pro, 2026-05-02.
---
Research conducted by SCOUT (Perplexity sonar-pro, 10 queries). Total Perplexity queries: 12. No WebSearch or WebFetch fallback required — all primary literature returned with citations. Memo word count: ~3,800.