# Caverio Companion — Full Product & UX Feedback for V2 **Product reviewed** - Live companion: https://caverio-companion.pages.dev/ - V1 wireframes: https://caverio-companion.pages.dev/wireframes-v1/ **Purpose of this document** This is not a cosmetic design review. It is a product-architecture and UX review intended to guide the next iteration of Caverio. The research foundation is already strong. The main V2 task is to convert that research into a product that feels **simpler, faster, more differentiated, and more obviously useful during an actual trading workflow**. The central recommendation is: > **Do not build Caverio as a richer crypto dashboard. Build it as a live situation model for a token.** The strongest product is already present inside the research: normalizing fundamentally different signals onto one clock, showing their sequence and relationship, tracking what changed, and preserving provenance/uncertainty. V2 should make that the product. --- # 1. Executive Summary Caverio has a genuinely differentiated idea, but the current wireframes still express too much of the **research taxonomy** directly in the interface. The research has many useful dimensions: - money flow - tracked wallets - traders - callers - attention - tape - holders - momentum - safety - narratives - peers - records - historical comparisons - data freshness - provenance - unknowns - profile logic - market context That is excellent for the analytical engine. It is too much as a first-order interface model. The core risk is that Caverio becomes a better-designed, more intelligent version of the same thing traders already have: a dense terminal with more panels. The V2 goal should instead be: > **A trader should understand the current situation in 5–10 seconds, then be able to drill into evidence.** A strong Caverio experience should answer five questions in order: 1. **What is happening?** 2. **Why does it matter now?** 3. **In what order did the important events happen?** 4. **What changed since I last looked / since I entered?** 5. **What evidence supports this interpretation, and what is still unknown?** Everything else is secondary. --- # 2. The Core Product Thesis The strongest idea in Caverio is not: > “One dashboard that replaces five crypto tools.” That is useful marketing language, but it is not the true moat. The stronger thesis is: > **Caverio reconstructs a token situation across multiple evidence sources on one shared timeline.** For example: - money begins arriving - a known or recurrent wallet enters - early holders retain or exit - caller attention begins later - public attention accelerates - liquidity changes - participant breadth expands - a previously bullish caller becomes neutral - a tracked trader reduces - the user enters - the situation later changes Most existing tools show these things in separate surfaces. Caverio can show the **relationship between them**. Examples of genuinely differentiated observations: - “Money arrived 32 minutes before attention.” - “Two tracked wallets entered before the first notable caller mention.” - “Attention is still rising, but early-holder retention has deteriorated.” - “Caller stance turned positive after price had already moved 118%.” - “Since you entered, one tracked participant reduced and early-holder exits accelerated.” - “Nothing material changed since you last checked.” - “The apparent signal is incomplete because holder coverage is stale.” This is much more valuable than merely adding another panel of analytics. --- # 3. The Product Category Caverio Should Aim to Own The product should feel less like: - a terminal - a screener - a wallet tracker - an alpha feed - a caller leaderboard - a market dashboard and more like: > **A live situation room for a token.** That phrase implies: - chronology - actors - evidence - changing conditions - uncertainty - state transitions - current interpretation - historical context - personal context This framing also gives the team a useful product test: > **Does this feature improve the user's understanding of the current token situation?** If the answer is no, the feature should probably be secondary, hidden, postponed, or removed. --- # 4. What Is Already Strong Several parts of the current thinking are excellent and should survive V2. ## 4.1 Multi-source chronology This is the strongest differentiator. The product should continue to align: - price - money - wallets - holder behavior - attention - callers - participant behavior - safety events - user entry/checkpoints onto one timeline. That timeline should become more important than any individual analytics panel. --- ## 4.2 “Since You Looked” This is one of the best ideas in the product. A returning user usually does not need the entire situation explained again. They need: > **What changed?** Examples: - second tracked wallet entered - first notable caller appeared - early-holder retention weakened - liquidity increased - attention accelerated - tracked trader reduced - safety state changed - no material change This is exactly how Caverio becomes a **companion** rather than another dashboard. It should be a major V2 primitive. --- ## 4.3 “Since You Entered” This may be even more valuable than “since you looked.” Once the user marks themselves as “in,” the product gains a personal reference point. Now the product can answer: > **Has the situation I entered materially changed?** That is an extremely powerful retained-use loop. The user does not need Caverio to tell them what to do. Caverio should tell them: - what changed - what stayed intact - what became weaker - what became stronger - what is unknown - when those changes happened relative to their entry That is useful without pretending to predict the future. --- ## 4.4 Provenance and uncertainty The product philosophy around provenance, stale data, gaps, and unknowns is very strong. Do not lose it. The opportunity is to make this **local and contextual**, rather than constantly exposing infrastructure telemetry. Examples: - “Holder data 11m old” - “X unavailable since 14:32” - “Wallet attribution confidence: medium” - hatched section in timeline - partial evidence label - “insufficient history” This makes the product trustworthy. --- ## 4.5 Adaptive situation profiles The research correctly recognizes that different token states require different emphasis. A 25-minute-old microcap and a 3-day-old $25M token should not have identical information hierarchy. That adaptive logic should remain. However, the user should not see implementation language such as: - Profile A - Profile B - Profile C The interface should simply adapt naturally. --- ## 4.6 Historical Record The Record concept is strategically important because it reinforces: > **Do not trust Caverio because it sounds confident. Inspect what it has observed.** That is a powerful brand position. Record should support: - historical situations - fixed-horizon outcomes - methodology - sample size - baselines - censored data - false positives - source limitations This is especially important because Caverio should avoid becoming another opaque “AI score” product. --- # 5. The Main V1 Problem: Research Taxonomy Leaking Into the Interface The current research model has many categories and internal concepts. V1 exposes too many of them directly. The Situation Room currently risks becoming a long sequence of: - market - safety - lights - chart - tape - money flow - holders - callers - participants - timeline - record - narrative - unknowns - freshness This creates two problems. ## 5.1 Cognitive cost The user has to reconstruct the situation themselves from many modules. That defeats part of the product’s purpose. Caverio should do more synthesis before presenting information. --- ## 5.2 Differentiation gets diluted The unique idea — multi-source chronology — becomes one module among many. That makes Caverio feel closer to a sophisticated terminal. The most unique part should instead dominate the product. --- # 6. V2 Design Principle: Synthesis First, Evidence Second Every important screen should have two layers. ## Layer 1 — Situation The system explains the current state in human-readable form. Example: **Money arrived first** 3 tracked wallets entered before notable public attention. Attention is now accelerating, while early holders remain mostly intact. Then show: - clock - key change - key uncertainty - what the system is waiting to learn --- ## Layer 2 — Evidence The user can inspect: - money flow - wallets - holders - callers - participant details - safety - provenance - methodology - historical record This should be drill-down material. The user should never need to read six separate panels just to reach the interpretation. --- # 7. Recommended Core Hierarchy Across the product, use a consistent six-level hierarchy. ## 1. STATE The clearest current interpretation. Examples: - Money arrived first - Attention only - Crowd arriving - Early holders weakening - Quiet / developing - Broad participation forming - Safety changed - Momentum cooling - No material change --- ## 2. EVIDENCE One or two sentences explaining why. Example: > 3 recurrent wallets entered before the first notable caller mention. Net tracked flow is +$42K. --- ## 3. TIME Show the relationship between events on the Caverio Clock. --- ## 4. CHANGE What changed since: - last look - watch - entry - previous alert - relevant checkpoint --- ## 5. WAITING FOR What missing evidence would materially change the interpretation. Examples: - broader wallet participation - early-holder stabilization - second independent attention source - fresh holder snapshot - confirmation that liquidity remains intact This is a very strong concept because it communicates uncertainty without forcing a prediction. --- ## 6. DETAILS Expandable evidence and analytics. This hierarchy should be used everywhere. --- # 8. The Caverio Clock Should Become the Signature Component The Clock should be the visual and conceptual center of the product. Right now, key inputs are separated into modules. Instead, Caverio should visually synchronize them. A conceptual example: ```text SPRNG $640K +212% MONEY ARRIVED FIRST 41m old 08:20 08:33 09:05 09:37 NOW PRICE ───────╱──────────╱──────────────────────── MONEY ● wallet ● wallet HOLDERS ↓ exits ATTENTION ▲ Pow ▲ Columbus SAFETY LP ✓ <── money led attention by 32m ──> SINCE YOU LOOKED + Second tracked wallet entered + Columbus turned bullish - Early-holder retention fell 72% → 53% WAITING FOR Broader independent participation ``` This is a product. A set of fourteen panels is a dashboard. --- # 9. The Clock Should Support Multiple Modes The same core timeline can adapt based on context. ## 9.1 Discovery mode Short compressed timeline. Focus on: - why now - latest important event - relationship between money and attention --- ## 9.2 Situation mode Full timeline. Possible lanes: - price - money - tracked wallets - holders - attention/callers - participant behavior - safety Not every lane has to be visible simultaneously. Allow intelligent collapsing. --- ## 9.3 Hold mode Timeline relative to the user's entry. Show: - entry marker - changes since entry - original thesis context - material deterioration/improvement - stance flips - participant exits --- ## 9.4 Historical mode Used in Record. Show frozen situations and subsequent outcome windows. --- # 10. Do Not Overload the Timeline The timeline can become cluttered very quickly. Use **semantic compression**. Instead of showing every raw event: > 37 mentions show: > Attention accelerated ×3 Instead of every wallet transaction: > 3 independent tracked wallets entered Allow expansion for the raw events. This keeps the main experience interpretive rather than forensic. --- # 11. Reconsider the “Lights” UX The eight-light model appears useful internally. Possible lights: - wallets - traders - callers - attention - tape - holders - momentum - safety The problem is that users may interpret: > more lights = better opportunity even if the research does not support that conclusion. That creates a misleading heuristic. ## Recommendation Keep lights as an **internal evidence model**, but do not make “X/8 lights” the dominant user-facing concept. Instead, convert combinations of evidence into descriptive states. Examples: ### MONEY FIRST Tracked/recurrent money precedes attention. ### ATTENTION ONLY Attention is accelerating without corresponding tracked participation. ### CROWD ARRIVING Money breadth and attention are rising together. ### EARLY MONEY LEAVING Attention remains elevated while early holders are reducing. ### DEVELOPING Insufficient evidence for a stronger state. ### SAFETY CHANGE A material contract/liquidity/deployer event occurred. The underlying lights can still be shown in a details drawer: > Evidence: Wallets ✓ Attention ✓ Holders ? Safety ✓ This is much easier to understand. --- # 12. State Language Must Be Descriptive, Not Predictive Avoid states that imply a recommendation. Bad: - Buy - Strong Buy - Bullish opportunity - Best setup - High conviction - Sell - Exit Better: - Money first - Attention accelerating - Holder retention weakening - Broad participation - Attention ahead of money - Quiet - Cooling - Safety changed The system should describe the observed situation. --- # 13. Explore — Core Product Question Explore should answer: > **Why should I look at this now?** Not: > How many metrics can I see before opening the token? The current Explore structure has several good elements, but it can be substantially simplified. --- # 14. Recommended Explore Structure ## Top ### Search / paste CA Always immediately accessible. ### Compact market weather One short line, not a dashboard. Example: > SOL active · <$1M strongest activity · runners cooling · tracked liquidity stable This provides context without turning Explore into Market Brief. --- ## Main sections Potential sections: ### Happening Now Tokens with meaningful new state changes. ### Newly Developing Early situations with incomplete evidence. ### Changed Since You Looked Tokens the user has previously opened where something material changed. ### Your Board Watched / entered tokens with recent state changes. ### Cooling Situations that were active but have lost momentum. Avoid too many concurrent categories. --- # 15. Explore Card / Row Design Every token result should answer: 1. What is happening? 2. Why now? 3. How old is that observation? 4. What is the next uncertainty? Example: ```text SPRNG $640K MONEY FIRST +212% 2nd tracked wallet entered 4m ago Money preceded notable attention by 31m ●●────────────▲ Waiting for: broader participation ``` Another: ```text AGRP $3.4M EARLY HOLDERS REDUCING 47% of first-50 buyers now out Attention continues to accelerate ●──────────▲▲▲ ↓↓↓ Changed 6m ago ``` This is more informative than an array of colored dots. --- # 16. Explore Should Not Require Decoding Avoid forcing the user to understand: - 8 light colors - 4 freshness states - profile labels - caller abbreviations - multiple score badges - unlabeled glyph systems A trader should be able to scan the page extremely quickly. State language does the heavy lifting. Visual encoding supports it. --- # 17. Situation Room — Recommended V2 Architecture The Situation Room should be the centerpiece of Caverio. The V1 version is analytically complete, but the V2 should be much more aggressively hierarchical. --- # 18. Situation Room — Above the Fold The user should initially see only the information needed to understand the situation. Recommended structure: ## Header - token - chain - age - market cap - price change - liquidity - search/copy/share/watch controls - user state: Watching / I'm In --- ## Current State Large clear label: > **MONEY ARRIVED FIRST** Supporting sentence: > 3 recurrent wallets entered before notable attention. Public attention is now accelerating. --- ## Caverio Clock The signature visualization. --- ## Since You Looked Maximum 3–5 meaningful items. Example: - + second tracked wallet entered - + Columbus changed stance to bullish - – early-holder retention fell 72% → 53% If nothing changed: > No material change since 09:14. This is valuable information. --- ## Waiting For Example: > Broader independent wallet participation. or: > Fresh holder snapshot — current holder data is 18m old. This communicates uncertainty and directs attention without recommending an action. --- # 19. Situation Room — Below the Fold Then allow the user to inspect evidence. Suggested order: 1. Money & participants 2. Holders 3. Attention & callers 4. Safety 5. Narrative / peers 6. Historical record 7. Provenance / methodology Do not give each section equal visual weight. --- # 20. Money & Participants This section should answer: - Is money entering or leaving? - Is it one actor or multiple independent actors? - Are participants recurrent / known? - Did they arrive before or after attention? - Are they still holding? - Has anyone materially reduced? Prioritize relationships over raw numbers. Good: > 3 independent recurrent wallets entered over 19m. First entry preceded attention by 31m. Less useful as primary UI: > 0x9f... bought $7,321 at 08:42:16 The transaction-level detail can still exist on expansion. --- # 21. Holders The Holders panel should emphasize trajectory. Useful dimensions: - early-holder retention - first-N buyer exits - concentration change - deployer/team movement - cohort behavior over time - new-holder breadth - notable cohort exits Prefer: > First-50 buyer retention: 72% → 53% in 38m over: > Holders: 1,382 Absolute holder count is secondary. --- # 22. Attention & Callers The strongest value is not simply: > 9 callers mentioned this. It is: - who spoke - when - relative to price/money - whether stance changed - whether attention is independent or copied - whether attention is expanding after a move Examples: > First notable caller mention occurred 31m after tracked money entered. > 4 caller mentions followed PowsGem within 12m. > PowsGem stance: bullish → neutral. > Public attention accelerated after +96% price move. These are much more useful. --- # 23. Caller Stance Changes Are Especially Valuable A static caller label has limited value. A **stance change** is much more meaningful. Examples: - bullish → neutral - bullish → stopped mentioning - neutral → bullish - entered → reduced - reiterated after drawdown The product should highlight state transitions. --- # 24. Safety Safety is necessary, but it should not dominate the experience unless something changes. Default state: > Safety intact Expand for: - liquidity - mint authority - deployer movement - contract flags - bundle issues - ownership - taxes - suspicious changes If a material event occurs, then Safety can become a primary situation state: > **SAFETY CHANGED** --- # 25. Unknowns “Unknown” is not a failure. It is part of the product. However, avoid a permanent large “Unknowns” panel. Instead surface unknowns where they matter. Examples: > Holder cohort unavailable for 11m. > Caller coverage incomplete. > Wallet identity confidence: low. > X attention unavailable. Then optionally provide one consolidated: > Data notes drawer. --- # 26. Data Health — Reduce Infrastructure Telemetry The current global data-health strip is conceptually honest, but too infrastructure-like. Showing all source heartbeat states on every screen makes the product feel like: > an internal observability dashboard rather than: > a trader-facing intelligence product. ## Recommendation Use a compact global state: > **Data health ● 2 issues** Then show local freshness exactly where relevant. Examples: - Price · 22s - Holders · 8m - Callers · 3m - X unavailable - BSC degraded This preserves trust with much less visual noise. --- # 27. Freshness Should Affect Visual Confidence Potential visual rules: ### Fresh Normal display. ### Aging Subtle timestamp. ### Stale Muted / dotted / explicit warning. ### Missing Hatched region or gap. ### Recovered after gap Show discontinuity. This is better than pretending the timeline is continuous. --- # 28. Remove “Profile A / Profile B / Profile C” From User-Facing UI Adaptive layouts are smart. Exposing the routing label is not useful. The product should simply change emphasis. ## First-hour token Prioritize: - first buyers - bundles - recurrent wallets - initial money - initial attention - liquidity - deployer safety ## Day-2 token Prioritize: - retention - participant changes - attention trajectory - recurrent money - stance shifts - narrative persistence ## Established token Prioritize: - broader holder behavior - sustained flows - peer/narrative context - participant persistence - market relationship The user should never have to care which “profile” was selected. --- # 29. Hold View / “I’m In” — Potentially the Stickiest Feature This should be treated as a primary product surface. Once a user enters, Caverio can become a monitoring companion. The core question changes from: > What is happening? to: > **Has the situation I entered changed?** That is an excellent retained-use loop. --- # 30. Hold View — Recommended Structure ## Header - token - entry time - optional entry price - current price - time held - original situation state Example: > Entered 2h 18m ago during **MONEY FIRST** --- ## Material Changes Since Entry Do not mix changed and unchanged states in one long severity list. Use: ### CHANGED - Early-holder retention: 72% → 47% - PowsGem: bullish → neutral - ryu reduced position - net tracked flow slowed ### STILL INTACT - deployer has not sold - liquidity remains intact - tracked money remains net positive ### UNKNOWN / DEGRADED - X unavailable - holder snapshot 16m old This structure is much easier to reason from. --- # 31. Hold View Should Preserve the Entry Context When the user says “I'm In,” save a snapshot of: - state - major evidence - clock - important participants - early-holder status - attention status - safety status - uncertainties Then compare current state against that snapshot. This makes “since entry” meaningful. --- # 32. Introduce the Concept of a Situation Diff This could become an important reusable internal object. A **Situation Diff** compares two timestamps. Possible comparisons: - now vs last look - now vs watch - now vs entry - now vs alert - now vs 1h ago The diff should classify: - added evidence - removed evidence - strengthening - weakening - unchanged - unknown due to stale/missing data This can power multiple surfaces. --- # 33. My Board Should Become a Core Top-Level Destination Instead of the product being mainly: > Explore → token → leave create an explicit retained workflow. Suggested tabs: ### Watching Tokens the user is monitoring. ### I'm In Tokens the user holds / has entered. ### Changed Everything with meaningful state changes since last look. ### Archived Past situations. My Board should be optimized for **delta**, not static metrics. --- # 34. My Board Row Design Example: ```text SPRNG Entered 2h 18m ago 2 material changes - Early-holder retention weakened - PowsGem turned neutral Still intact - Net tracked flow positive - Safety unchanged ``` The user should not need to reopen every token just to discover nothing changed. --- # 35. Notifications Should Eventually Be State-Change Based Avoid noisy alerts like: - price up 5% - mention detected - wallet transaction happened Prefer material changes: - second independent tracked wallet entered - tracked participant reduced - early-holder retention crossed meaningful change threshold - caller stance changed - safety changed - attention accelerated without money - data coverage became unreliable The notification should explain **why the event matters in the situation**. --- # 36. Market Brief — Useful, But Secondary The Market Brief has value as context. However, this is not where Caverio is most differentiated. A full page containing: - stablecoin supply - BTC dominance - ETF flows - chain heat - cap-band heat - broad activity - terminal-level metrics can quickly make the product feel like generic crypto analytics. ## Recommendation Keep a lightweight “market weather” layer in the primary experience. Example: > SOL hot · sub-$1M active · runner follow-through weakening Then retain the full Market Brief as a secondary page for users who want it. Do not allow it to consume disproportionate early development effort. --- # 37. Market Weather Should Be Descriptive Avoid turning macro context into a pseudo-predictive banner. Good: > Sub-$1M Solana activity is elevated versus the last 7 days. Bad: > Bullish environment for microcaps. Keep it observational. --- # 38. Zones — Weakest Conceptual Fit The “buy / hold / sell / past” zone concept conflicts with the strongest Caverio philosophy. Even if labeled “descriptive, not instructional,” the semantics still look like trading advice. More importantly, historical pattern evidence can easily be overinterpreted. ## Recommendation Replace **Zones** with something like: ### Similar Situations Example: > 54 historical situations had a comparable pattern. Then show: - distribution of subsequent outcomes - time to event - drawdown before expansion - failure cases - sample size - methodology - censoring - confidence limitations Example: ```text 54 comparable situations 29 reached +20% before -30% 10 reached -30% first 15 remained inside the band during the observation window Median time to +20% where achieved: 1h 40m ``` This is more aligned with the product's philosophy. It gives context without converting observations into commands. --- # 39. Do Not Compress Historical Evidence Into One Score Avoid: > Historical setup score: 8.3/10 Prefer the raw contextual distribution. The product's trust advantage is transparency. Do not throw that away for convenience. --- # 40. Record — Important but Not Primary Navigation Record is valuable for trust, research, and power users. It should expose: - frozen historical situations - evidence available at the time - what occurred later - caller outcomes - state frequencies - false positives - comparisons against random baselines - data limitations - methodology changes However, the daily product loop should probably be: > **Explore → Situation → Watch / I'm In → Return → Changes** Record supports that loop. It is not the loop itself. Possible placement: - Evidence - Record - Research - Method as a secondary destination. --- # 41. Recommended Top-Level Information Architecture A cleaner V2 architecture: ## DISCOVER What deserves attention now? Contains: - search - happening now - newly developing - market weather - changed situations --- ## SITUATION The token's current situation. Contains: - state - evidence - clock - since-you-looked - waiting-for - drill-down evidence --- ## MY BOARD What changed in things I care about? Contains: - Watching - I'm In - Changed - Archived --- ## SECONDARY - Market Weather / Brief - Record / Evidence - Search - Settings This is easier to understand than many equal-priority destinations. --- # 42. The Primary Product Loop Design the entire product around this loop: ## 1. Discover User sees: > Why now? ## 2. Open Situation User understands: > What is actually happening? ## 3. Watch or Enter User declares interest/state. ## 4. Leave Caverio continues collecting evidence. ## 5. Return Caverio says: > Here is what changed. ## 6. Reassess User inspects the relevant evidence. This loop should feel dramatically faster than checking: - DexScreener - wallet trackers - Telegram - X - holder tools - caller feeds individually. --- # 43. “Waiting For” Is an Excellent Product Primitive This deserves more emphasis. Instead of pretending certainty, Caverio can explicitly say what evidence is incomplete. Examples: > Waiting for broader wallet participation. > Waiting for fresh holder data. > Waiting for an independent attention source. > Waiting for evidence that first buyers are retaining. > Waiting for sufficient history. This does several things: - communicates uncertainty - creates a reason to return - avoids false confidence - teaches the user how Caverio interprets evidence - makes future alerts meaningful --- # 44. “Why This State?” Should Always Be Inspectable Every synthesized state needs transparent reasoning. Example: ### MONEY FIRST **Why this state** - recurrent wallet A entered at 08:31 - recurrent wallet B entered at 08:42 - tracked net flow +$42K - first notable caller mention at 09:03 - attention acceleration began at 09:07 This makes Caverio understandable rather than magical. --- # 45. Avoid Opaque AI Language Do not rely on: - AI confidence - AI conviction - AI opportunity score - AI trade score If machine reasoning is used internally, present the result as evidence-backed synthesis. The strongest Caverio brand is: > observable, inspectable intelligence not: > trust our AI. --- # 46. Design Language — Desired Character The UI should feel: - analytical - calm - precise - alive - dense but not crowded - high-trust - event-oriented - dark-terminal adjacent without becoming cliché - modern rather than gamer-like Avoid: - casino aesthetics - neon overload - constant red/green urgency - oversized numeric dashboards - too many glowing badges - generic “crypto terminal” visual language The differentiator should come from **information design**, not decoration. --- # 47. Motion Can Be Functional Use motion carefully to communicate changes. Examples: - new clock event softly enters - “since you looked” diff animates once - lane expands when event is selected - timeline jumps to entry marker - stale data subtly changes treatment - state transition briefly highlights Do not use constant motion. The product should feel live, not restless. --- # 48. Color Should Represent Meaning, Not Decoration Suggested semantic families: - positive evidence / increase - weakening / deterioration - neutral information - unknown - stale - safety / critical - user-specific marker Avoid using green = buy and red = sell. Because the product is descriptive, color should describe **change or evidence quality**, not recommended action. --- # 49. Make Event Relationships Interactive Clicking an event should reveal: - exact timestamp - source - relevant entity - provenance - related event - before/after context Example: > “Money led attention by 31m” Clicking it can highlight: - first tracked money event - first notable attention event - the interval between them This is where Caverio can feel unique. --- # 50. Event Types Should Have a Consistent Grammar Define a small event language. Possible types: ### Money - entered - added - reduced - exited - net flow changed - breadth changed ### Holder - retention improved - retention weakened - concentration increased - cohort exited ### Attention - first notable mention - acceleration - slowdown - stance change - independent source joined ### Safety - liquidity changed - deployer moved - authority changed - flagged condition appeared ### User - watched - entered - checked - alert received The same grammar should be used across Clock, Explore, My Board, and notifications. --- # 51. Entity Pages May Be Useful Later Potential future entity pages: - wallet - caller - tracked trader - deployer - narrative But do not let these distract from the situation model in MVP. Entity intelligence should first exist in context. Example: > Wallet `ryu` has entered 7 previously observed situations. Then allow deeper navigation later. --- # 52. Narrative Intelligence Should Remain Secondary Initially Narratives can be useful for larger/older tokens. But they risk introducing another giant research domain. For V2/MVP: Use narrative primarily to answer: > What broader story is this token currently associated with? and: > Is that narrative itself gaining or losing attention? Do not overbuild narrative scoring before the core Situation loop proves sticky. --- # 53. Peer Context Should Be Situational Do not create a generic “Peers” table just because comparable tokens exist. Use peers when they help explain the situation. Example: > 4 of 7 similar Solana AI-agent tokens are seeing attention acceleration today. or: > This token is the only one in the peer set with rising tracked wallet breadth. That is useful contextual evidence. --- # 54. The MVP Should Be Cut Harder The full product vision is strong, but trying to ship all research surfaces at once risks obscuring the core hypothesis. The MVP should test: > **Will traders repeatedly reopen Caverio because it tells them something they cannot easily see in their existing workflow?** Recommended MVP: ## Explore - ranked situations - descriptive state - why now - compact clock - waiting-for - search ## Situation - state - evidence - full Clock - money - holders - attention - safety - since-you-looked - provenance ## My Board / Hold - Watching - I'm In - situation diff - since-entry changes ## Record Minimal methodology/history proof. --- # 55. Defer or Minimize for MVP Potential candidates for later waves: - full Market Brief - broad macro analytics - complex narrative engine - deep peer engine - Zones - advanced historical similarity engine - every chain - every caller ecosystem - elaborate scorecards - extensive trader profile pages - highly granular alert configuration The MVP should prove the core loop first. --- # 56. Platform Dependency Risk Caller / Telegram intelligence may be highly valuable. However, avoid building a product whose core usefulness collapses if one content source becomes unavailable. The core should remain useful with: - on-chain tape - wallet recurrence - money flow - holder behavior - liquidity / safety - available attention sources - public social data where permitted - chronology - user checkpoints Caller intelligence should be a powerful additional lane, not the structural foundation. --- # 57. Build a Source-Abstraction Layer Internally, event objects should not depend too tightly on one provider. Example event schema concept: ```text event_id token_id event_type event_subtype timestamp source_type source_id entity_id value_before value_after confidence freshness provenance metadata ``` This lets the product survive provider changes. --- # 58. Model “Confidence” as Data Quality, Not Predictive Confidence Useful: > Wallet attribution confidence: medium. > Attention coverage: partial. > Holder freshness: low. Less useful: > Trade confidence: 82%. Confidence should describe **how trustworthy the observation is**, not how likely price is to rise. --- # 59. Monetization — Validate Substitution, Not Just Satisfaction A trader saying: > “This looks cool” does not prove willingness to pay. The better question is whether Caverio replaces meaningful work. During beta, periodically ask: ### Before Caverio, what would you have opened to answer this? - DexScreener - GMGN - wallet tracker - Telegram - X - holder tool - multiple of the above - I would not have checked And: ### What did Caverio change? - noticed something earlier - found a token - avoided entering - entered with better context - monitored a position more easily - reduced time spent checking tools - changed nothing This measures product value more directly. --- # 60. Stronger Willingness-to-Pay Signal Ask: > If Caverio disappeared tomorrow, what would you do instead? High-value answers: - “Go back to 4–5 tools.” - “I’d miss the change tracking.” - “I’d have no replacement for the timeline.” - “I’d need to manually monitor Telegram + wallets again.” Weak answer: - “Probably just use DexScreener.” This will reveal whether the differentiated workflow is real. --- # 61. Pricing Should Follow Habit Formation Do not optimize pricing before confirming: - weekly retained usage - repeat token monitoring - My Board usage - return sessions triggered by changes - substitution of existing tools - high-value features users would miss Once those are clear, pricing becomes much easier to justify. --- # 62. Product Metrics That Matter Do not rely primarily on: - total page views - token views - session duration Track product-loop metrics. ## Discovery - Explore → Situation open rate - search vs ranked discovery - percent of opened situations watched ## Monitoring - watch rate - I'm In rate - return-to-token rate - My Board revisit frequency ## Change utility - percent of return visits where a Situation Diff exists - clicks on “since you looked” - clicks on “since entry” - alert → Situation open rate ## Substitution - reported tools replaced - number of external tools still needed - self-reported time saved ## Trust - provenance expansion rate - Record usage - stale-data warning interactions - user corrections / disputed observations --- # 63. Suggested V2 Navigation Possible primary navigation: ```text CAVERIO Discover My Board Search ────────── Market Record Settings ``` When inside a token, the Situation Room becomes the dominant page. Do not make every research domain a top-level navigation destination. --- # 64. Suggested Situation Header Example: ```text SPRNG / SOL $642K +212% 41m old MONEY ARRIVED FIRST 3 recurrent wallets entered before notable attention. Attention is now accelerating. [ Watch ] [ I'm In ] ``` The title should communicate the situation, not just token stats. --- # 65. Suggested Clock Layout Desktop: ```text 08:20 08:33 09:05 09:37 NOW PRICE ──────────╱──────────╱───────────── MONEY ● A ● B HOLDERS ↓ exits ATTENTION ▲ Pow ▲ Col SAFETY ✓ LP USER │ last look ``` Below: > Money preceded notable attention by 32m. On mobile, collapse to 3–4 lanes with horizontal scroll or event summary. --- # 66. Suggested “Since You Looked” Module ```text SINCE YOU LOOKED · 38m + Second recurrent wallet entered 12m ago + Columbus changed stance: neutral → bullish 9m ago – First-50 buyer retention fell 72% → 53% 4m ago ``` Rules: - maximum 3–5 default events - sort by materiality, not raw chronology - allow “View all” - show “No material changes” when appropriate --- # 67. Suggested “Waiting For” Module ```text WAITING FOR Broader independent wallet participation. Current tracked buying is concentrated in 3 related/recurrent participants. ``` or: ```text WAITING FOR Fresh holder data. The most recent holder cohort snapshot is 18m old. ``` This should be simple and explicit. --- # 68. Suggested Evidence Drawer ```text WHY THIS STATE ✓ 3 recurrent wallets entered ✓ first tracked buy preceded attention by 31m ✓ tracked net flow +$42K ✓ attention acceleration began 8m ago ? early-holder data is 11m old [View raw events] ``` This keeps synthesis transparent. --- # 69. Suggested Hold View ```text YOU ENTERED 09:14 · $0.0042 State at entry: MONEY FIRST CURRENTLY Early-holder retention is weakening. Other entry conditions remain mostly intact. CHANGED - Early-holder retention 72% → 47% - PowsGem bullish → neutral - ryu reduced STILL INTACT - tracked net flow positive - liquidity unchanged - deployer has not sold UNKNOWN - X unavailable for 22m ``` No explicit instruction is needed. The information is already useful. --- # 70. Suggested Explore Row ```text SPRNG $640K +212% MONEY FIRST 2nd recurrent wallet entered 4m ago Money led attention by 31m ●●──────────▲ Waiting for broader participation ``` Use hover/expand for additional evidence. --- # 71. Suggested Compact Market Weather ```text MARKET WEATHER SOL activity elevated Sub-$1M launches active Runner follow-through cooling ``` Allow click-through to full context. --- # 72. Suggested Record Entry ```text SPRNG · frozen at 09:14 State: MONEY FIRST Evidence available at the time: - 2 recurrent wallets - attention lag: 31m - early-holder retention: 74% - liquidity intact Subsequent path: +20% reached in 46m -30% not reached during observation window [Open frozen situation] ``` The key is preserving what was known **at that time**. Do not backfill future information into the historical view. --- # 73. Frozen Historical Rooms Are a Powerful Trust Mechanism This concept deserves emphasis. A user should be able to inspect: > What did Caverio actually know at 09:14? not: > What does the system now know about what happened? That prevents hindsight bias. It also enables honest evaluation. --- # 74. Preserve Data Gaps in Historical Rooms If data was missing at that time, show it. Do not retrospectively make historical situations look cleaner than they were. Examples: - caller stream unavailable - holder snapshot stale - wallet attribution unresolved - price gap Trust compounds when the product shows its limitations. --- # 75. Avoid Metric Theater Every visible metric should justify itself. Ask: > What decision or understanding does this metric improve? If the answer is unclear, demote it. Potential metric theater examples: - too many top-level counters - giant total-holder numbers with no trajectory - raw mention counts with no timing - generic sentiment percentages - composite scores - multiple redundant momentum badges Caverio should prioritize **relationship metrics**. --- # 76. Relationship Metrics Are the Real Gold Examples: - money → attention lag - attention → price lag - early holders → attention divergence - participant breadth change - caller stance → subsequent participation - user entry → evidence changes - safety event → liquidity reaction - first buyer cohort → retention trajectory These are more differentiated than isolated values. --- # 77. Think in State Transitions, Not Static Snapshots Static: > Attention: high Better: > Attention accelerated 3× over 18m. Static: > Holder retention: 53% Better: > Early-holder retention fell 72% → 53% since last look. Static: > 3 tracked wallets Better: > 2 new independent tracked wallets entered after your last check. This should become a universal product principle. --- # 78. Product Copy Should Be Extremely Precise Avoid vague copy: - strong activity - interesting - bullish signals - smart money - high momentum Prefer observable copy: - 3 recurrent wallets entered - net tracked flow turned positive - attention rate doubled - 47% of first-50 buyers have exited - 2 notable callers changed stance The system should sound like an analyst, not a promoter. --- # 79. “Smart Money” Should Be Used Carefully If a wallet is designated as “smart,” users may assume predictive value. Prefer more descriptive entity labels: - recurrent wallet - tracked wallet - historically early participant - high-frequency participant - known caller-linked wallet - previously observed trader Then provide historical context if available. --- # 80. Avoid Ranking People With a Single Magic Score Caller/trader pages should show dimensions, not one opaque score. Possible dimensions: - sample size - median timing - frequency - early vs late participation - observed follow-through - drawdown profile - recurrence - stance-change behavior This matches Caverio’s transparency positioning. --- # 81. Search Can Be a Major Entry Point Support: - token ticker - contract address - wallet - caller - trader/entity Eventually: > “show me tokens where money arrived before attention” could become a natural-language query. But do not overbuild this before the core token situation experience. --- # 82. Consider a Global “Changes” Feed A very useful future surface: > **Changes** Across watched / entered situations: ```text SPRNG Early-holder retention weakened 4m ago ANON 2nd tracked wallet entered 7m ago PEPE2 Safety changed: liquidity reduced 11m ago ``` This is more useful than a generic notification center. It is effectively a personal intelligence feed. --- # 83. Avoid Turning Changes Into Noise Only emit a change when it crosses a materiality threshold. Internal logic can consider: - magnitude - novelty - independence - time since previous event - user relevance - state-transition significance Do not show every micro-change. --- # 84. Materiality Should Be Explainable If Caverio decides something is “material,” allow inspection. Example: > Marked material because early-holder retention changed by 19 percentage points within 14m. This will help debugging and user trust. --- # 85. Recommended Product Principles The agent should use these whenever making an unspecified design decision. ## Principle 1 — Situation before metrics Explain first. Measure second. ## Principle 2 — Relationships before isolated values Sequence and divergence matter more than totals. ## Principle 3 — Change before repetition Returning users care about what changed. ## Principle 4 — Evidence before confidence Show the basis, not a magic score. ## Principle 5 — Unknown is a valid state Never hide uncertainty. ## Principle 6 — Local freshness over global telemetry Put data quality where it affects interpretation. ## Principle 7 — Personal checkpoints matter Last look, watch time, and entry time create valuable context. ## Principle 8 — Fewer states, clearer language Do not create dozens of jargon-heavy classifications. ## Principle 9 — Adaptive depth Young microcaps, mature tokens, and holdings require different emphasis. ## Principle 10 — The Clock is the spine Every major analytical dimension should ultimately connect to time. --- # 86. Concrete V2 Component Set A possible component architecture: ## Global - `SearchBar` - `MarketWeatherStrip` - `DataHealthIndicator` - `Nav` - `EntityPopover` ## Situation synthesis - `SituationState` - `SituationSummary` - `WhyThisState` - `WaitingFor` - `SituationDiff` - `StateTransitionBadge` ## Timeline - `CaverioClock` - `ClockLane` - `ClockEvent` - `ClockRelationship` - `GapIndicator` - `UserCheckpoint` ## Evidence - `MoneyEvidence` - `HolderEvidence` - `AttentionEvidence` - `ParticipantEvidence` - `SafetyEvidence` - `NarrativeEvidence` - `PeerEvidence` ## Personal - `WatchButton` - `EntryMarker` - `HoldSummary` - `ChangedSinceEntry` - `StillIntact` - `UnknownSinceEntry` ## History - `FrozenSituation` - `HistoricalComparison` - `MethodologyDrawer` - `SampleSizeCaption` This component vocabulary is closer to the actual product thesis. --- # 87. Suggested Data Model: Situation State Conceptual object: ```json { "state": "money_first", "headline": "Money arrived first", "summary": "3 recurrent wallets entered before notable attention.", "evidence_ids": ["e1", "e2", "e3"], "started_at": "...", "last_changed_at": "...", "waiting_for": [ { "type": "participation_breadth", "text": "Broader independent wallet participation" } ], "data_quality": "good" } ``` The frontend should consume a synthesized state object rather than independently trying to infer the narrative from raw widgets. --- # 88. Suggested Data Model: Situation Diff ```json { "from": "last_view", "from_timestamp": "...", "to_timestamp": "...", "material_changes": [ { "category": "holders", "direction": "weakened", "headline": "Early-holder retention fell", "before": 0.72, "after": 0.53, "timestamp": "..." } ], "unchanged_conditions": [], "unknown_conditions": [] } ``` This can power: - Since You Looked - Since Entry - alerts - My Board - email/push summaries later --- # 89. Suggested Data Model: Relationship ```json { "type": "lead_lag", "source_event": "tracked_money_entry", "target_event": "attention_acceleration", "duration_seconds": 1920, "display": "Money led attention by 32m" } ``` Relationships should be first-class objects, not frontend-only labels. This is a major potential moat. --- # 90. V2 Prioritization ## P0 — Must define before visual polish 1. Situation state taxonomy 2. Material event taxonomy 3. Situation Diff rules 4. Clock event model 5. relationship/lead-lag model 6. user checkpoints 7. freshness / unknown logic If these are unclear, redesigning cards will not solve the product problem. --- ## P1 — Core V2 screens 1. Explore 2. Situation Room 3. My Board 4. Hold / I'm In state 5. Search These should be built around the new hierarchy. --- ## P2 — Trust layer 1. provenance 2. data health 3. frozen rooms 4. methodology 5. Record --- ## P3 — Broader intelligence 1. Market Brief 2. narratives 3. peers 4. historical similarity 5. entity pages --- ## P4 — Advanced workflows 1. notifications 2. user-configurable alerting 3. API 4. team/shared watchlists 5. research export 6. natural-language query --- # 91. Suggested First V2 Prototype Scope Do not redesign the entire platform at once. Prototype these three screens together: ## Screen A — Explore Contains: - market weather - 6–10 token situations - clear state - why now - mini Clock - changed-since-last-look indication --- ## Screen B — Situation Room Contains: - hero state - summary - full Clock - Since You Looked - Waiting For - 4 evidence sections --- ## Screen C — Hold Contains: - entry state - current state - Changed - Still Intact - Unknown - entry-relative Clock If these three screens feel coherent, the product has a strong spine. --- # 92. What to Remove From the Next Prototype For clarity, deliberately remove or hide: - visible Profile A/B/C labels - full permanent data heartbeat strip - 8-light count as primary status - full Market Brief from primary flow - Zones - giant unknowns section - raw tables above the fold - redundant token metrics - every evidence section being expanded by default - multiple competing scores This will make the differentiated product easier to see. --- # 93. What to Make Bigger in the Next Prototype Increase emphasis on: - Situation State - Caverio Clock - Since You Looked - Since Entry - relationships / lead-lag - material changes - Waiting For - provenance when needed These should visually define Caverio. --- # 94. Success Criteria for the V2 Prototype When a user opens a Situation Room, they should be able to answer within 5–10 seconds: 1. What is happening? 2. What changed? 3. What happened first? 4. Is the situation strengthening, weakening, or simply changing? 5. What evidence supports that? 6. What is still unknown? If the user must scroll through several cards to answer those questions, V2 is still too dashboard-like. --- # 95. User Testing Tasks Do not ask: > “Do you like this UI?” Give specific tasks. ### Task 1 > You saw this token 40 minutes ago. Tell me what changed. ### Task 2 > Did money or attention arrive first? ### Task 3 > You entered two hours ago. What has changed since entry? ### Task 4 > What evidence is missing or stale? ### Task 5 > Why does Caverio currently call this “Money First”? ### Task 6 > What would you check next if Caverio did not exist? Measure: - time to answer - errors - scrolling - external tool instinct - terminology confusion --- # 96. Terminology to Validate Potential product terms: - Situation - State - Clock - Changed - Since You Looked - Since Entry - Waiting For - Evidence - Record - Market Weather - My Board These are intuitive. Terms to be cautious with: - lights - profile - conviction - zone - smart money - alpha - score - AI confidence The latter group carries assumptions the product may not want. --- # 97. Brand Positioning Opportunity Caverio can occupy a strong philosophical position: > **We do not predict the token. We reconstruct the situation.** or: > **See what happened, in what order, and what changed.** or: > **A live situation room for every token.** The final marketing language can evolve, but the product should embody that idea. --- # 98. Why This Position Is Defensible Many competitors can add: - another score - wallet labels - caller feeds - social sentiment - holder charts - alerts It is harder to build a coherent system that: 1. normalizes these signals 2. places them on one clock 3. understands state transitions 4. preserves data provenance 5. tracks user-relative changes 6. maintains historical frozen views That system becomes more useful as its history grows. --- # 99. Potential Long-Term Moat If executed properly, Caverio's moat may become its **event graph + historical situation archive**, not the UI itself. Over time it can accumulate: - token events - participant histories - stance transitions - relationship timings - holder cohort behavior - situation state transitions - user-relative diffs - historical outcomes That dataset can support progressively better: - context - comparison - anomaly detection - historical analogs - participant characterization without forcing the product into opaque prediction. --- # 100. Final Recommendation The research behind Caverio is already sophisticated enough. The next phase should not add more analytical dimensions. It should **compress the intelligence into a clearer product.** The most important move is: > **Turn the common clock from one feature into the spine of Caverio.** Then organize the experience around: 1. **STATE** 2. **EVIDENCE** 3. **TIME** 4. **CHANGE** 5. **WAITING FOR** 6. **DETAILS** The combination of: - **Caverio Clock** - **Situation State** - **Since You Looked** - **Since You Entered** - **Waiting For** - **transparent provenance** is where the product feels meaningfully different. Everything else should orbit those concepts. The V2 design should feel **simpler than the V1 despite Caverio knowing more**. That is the standard to aim for. --- # 101. Agent Action Checklist ## Product architecture - [ ] Reframe product around “live token situation,” not “multi-tool dashboard” - [ ] Define small user-facing Situation State taxonomy - [ ] Define event taxonomy - [ ] Define relationship / lead-lag objects - [ ] Define Situation Diff - [ ] Define user checkpoints: viewed, watched, entered - [ ] Define materiality logic - [ ] Define stale/unknown states ## Explore - [ ] Make “Why now?” the primary row/card logic - [ ] Replace dominant 8-light UI with descriptive states - [ ] Add compact Clock/relationship visualization - [ ] Add Changed Since You Looked - [ ] Add Waiting For - [ ] Reduce raw metrics ## Situation Room - [ ] Hero the Situation State - [ ] Hero the Caverio Clock - [ ] Put Since You Looked above the fold - [ ] Add Waiting For - [ ] Move detailed evidence below the fold - [ ] Convert raw modules into evidence sections - [ ] Preserve provenance - [ ] Make data freshness local ## Hold / My Board - [ ] Save entry snapshot - [ ] Compare now vs entry - [ ] Split Changed / Still Intact / Unknown - [ ] Add entry marker to Clock - [ ] Make My Board delta-first - [ ] Show “No material change” explicitly ## Trust - [ ] Keep frozen historical rooms - [ ] Preserve historical data gaps - [ ] Expose methodology - [ ] Avoid opaque prediction score - [ ] Avoid hidden backfilled evidence ## Scope - [ ] De-prioritize full Market Brief - [ ] De-prioritize Zones - [ ] De-prioritize extensive narrative/peer systems - [ ] Keep caller intelligence modular - [ ] Prove Explore → Situation → Monitor → Return loop first --- # 102. Short Version for the Build Team If the team remembers only one thing: > **V1 contains a lot of intelligence, but makes the user assemble the situation. V2 should assemble the situation first, then let the user inspect the intelligence.** Caverio should not win because it has more panels. It should win because, when five different things are happening at once, it tells the user: > **what happened, in what order, what changed, and what is still unknown.**