Synthesis: the five-pillar architecture

Part of the «Waze для рыбалки» wiki · Status: v1, 2026-07-02
Full Russian version: Рыбалка_Waze_Исследование_2026-07-02.pdf, «Синтез» section
Related: Vision · Roadmap · Design principles

Eight research streams converge on one architecture:

1. Two data layers: non-rival public, spots gated

Public: ice/safety, water conditions, access, regulations, waterbody-level heatmaps, conditions forecast — gdebenz mechanics apply (one tap, confirmations, freshness decay) + Latvia's open data. Gated: spots — coarse default (waterbody/sector), delayed publication, private circles, give-to-get. Public exact pins: never.
→ Grounded in: gdebenz, secret-spots

2. Geolocation: verify exact, publish coarse (Piotr's principle)

GPS-verified reports get more weight and build author reputation — but the verified pin is never displayed. Verification is private (on-device capture, EXIF, mock-GPS detection, impossible-travel checks); display is always waterbody/sector. Precision buys trust, not spot-burning. Strictly opt-in — requiring GPS would trigger refusal or fake coordinates.
→ Grounded in: reputation-mechanics (iNaturalist obscuring, Strava k-anonymity lesson), Fishbrain 3-tier privacy

3. Reputation: corroboration for freshness, reputation for honesty

Anonymous confirmations suffice for cold start (gdebenz: 100k reports, zero accounts). But anglers lie strategically → reputation layer needed. Decouple fun leaderboard from trusted data layer (eBird model). Thank-you economy ("this tip worked") as primary reputation input. Private logbook as base reward. Anti-abuse at launch (gdebenz was gamed within days: 305 fake marks, 3 accounts, 1 hour).
→ Grounded in: reputation-mechanics, gdebenz

4. Hub: router over open data, not aggregator of apps

Integrate free & legal: LVĢMC open data (hourly water level/temp, 14-day hydro forecasts — CC0 APIs), fish-water map layers, free weather/solunar APIs, YouTube API, ribak.lv Atom feed, copeslietas.lv news RSS, Mana Cope deep links, Deeper user CSV exports. Don't touch: Facebook groups (API dead, scraping toxic — use in reverse as distribution), Fishbrain/FISHSURFING (walled), full forum ingestion (CV-Online v Melons precedent). Our exclusives where no API exists: live bite crowdsourcing, ice data (no machine-readable source exists), licensed-waters registry (official one delisted), paid-pond directory + booking (all book by phone).
→ Grounded in: hub-integrations, paid-ponds

5. Forecast: honest "conditions score", not "bite promise"

Build on species-specific water temperature (the one well-supported driver) + water level + diel rhythm. Solunar = debunked folklore (demote to labeled secondary); pressure = trend/front signal only, low weight. Frame: weather changes catchability, not abundance. Show the "why" behind every score. Latvia's hourly water data + 14-day hydro forecasts are an advantage no competitor uses.
→ Grounded in: forecasting