AI & Oracles

Domain: Machine intelligence, targeting, moderation, and content analysis Modules:

• Boost Targeting AI

• Content Similarity Embeddings

• SemanticTrustOracle

The AI & Oracles layer powers the platform’s intelligent content routing, moderation context, and post visibility logic. These systems are entirely governed by rules discussed and locked into the TRN ecosystem—there is no speculative personalization, no black-box behavior, and no deviation from deterministic expectations. AI is used to assist, not to control.

---

🎯 Role of AI Systems

AI on the TRN platform is used for:

• Boost Targeting: Finding the most resonant audiences for promoted posts

• Content Similarity Scoring: Identifying posts with similar themes, tone, or emotional payload

• Moderation Support: Providing embeddings and triggers for geo-restricted enforcement by country NFTs

• Semantic Auditing: Flagging edge cases, sensitive content, or illegal material for human/DAO oversight

• Prompt-Based Campaign Design: Allowing creators and advertisers to describe their goals in plain English and generate effective targeting prompts

All AI modules are deterministic, auditable, and governed by public logic and constraints wherever applicable. Only regionally restricted post-level bans are kept confidential for platform integrity.

---

🗃 AI Components Summary

Module	Purpose	User Facing?	Enforcement
Boost Targeting AI	Embedding search and targeting for ads	Yes	Yes (via ad delivery engine)
Content Similarity Embeddings	Clustering and semantic scoring	No	Yes (used for moderation + targeting)
SemanticTrustOracle	Publicly visible trust constraints and semantic categories	Yes	Yes (used by GeoOracle + DAO AI)

---

🔐 Enforcement Transparency

• Country NFTs can publish public category-level constraints

  • e.g. “No NSFW in Country X” or “No political memes in Region Y”

• The actual enforcement of post-level bans is hidden

  • Only revealed if content leaks via cross-region access

  • Ensures moderation integrity and prevents evasion

AI detects content violations using semantic triggers, text/image/audio embeddings, and regional constraints.

---

🧭 AI Usage by Role

Role	AI Access	Permissions
Users	View embeddings and boost reports	Can enable ads, disable targeting
Boosters	Use prompts, view targeting scope and reach	Cannot override AI targeting filters
Country NFT Holders	Publish constraints, manage semantic triggers	Cannot see private enforcement logs
DAO	Approves moderation rules and constraints	Oversees all AI modules except regional secrets

---

🚨 Manipulation Defense

• Boosters cannot manipulate reach by modifying prompts mid-campaign

• Embedding similarity scores are recorded and immutable during the boost

• All user actions (retrns, brns, subscriptions) are compared against AI bot detectors

• Detection includes repetition patterns, reaction timing, and semantic variance

• Boosted ads can trigger burn responses if users react poorly

  • Burn ratios impact campaign effectiveness and refund eligibility

---

🔄 Integration with Other Modules

AI insights are consumed by:

• BoostingModule → For targeting and payout logic

• GeoOracle → For constraint enforcement

• CountryRulesetManager → For semantic input exposure

• FlagEscalator → For fraud + illegal content escalation

• LottoModule → Resonance weighting and winner selection input

---

🔍 Related Modules

• Boost Targeting AI

• Content Similarity Embeddings

• SemanticTrustOracle

• GeoOracle

• CountryRulesetManager