Seshat

Team

Developers

hybernateAE

Founder · Lead Engineer

BaseDeFiTypeScriptCloudflareAI/ML

About

Builder focused on on-chain intelligence and developer tooling for the Base ecosystem. Seshat started as a personal tool to track unclaimed creator fees on Clanker tokens — it grew into a full self-learning scoring system.

The system runs entirely on Cloudflare's edge infrastructure: no servers, no databases to maintain, no ops overhead. Agentic AI loops powered by Claude research every project automatically — its authenticity, the real team behind it, and how its narrative compares to prior art. The scoring model updates itself every week using observed outcomes.

Stack

ComputeCloudflare Workers (TypeScript)
DatabaseCloudflare D1 (SQLite)
CacheCloudflare KV
StorageCloudflare R2
QueuesCloudflare Queues
VectorsCloudflare Vectorize + Workers AI (bge embeddings)
ChainBase mainnet (viem + Alchemy)
IndexingBitquery GraphQL · Alchemy webhooks
DataBankr · Dexscreener · CoinGecko · DeFiLlama · FrontrunPro
AIAnthropic Claude (agentic research, identity, swarm, chat)
FrontendNext.js 15 · Framer Motion · Tailwind

Build timeline

Phase 0

Concept — spotted unclaimed creator-fee opportunity on Bankr/Clanker

Phase 1

Fee scanner: Bankr API discovery, label-aware WETH extraction, V1/V2 ABI fallback

Phase 1

Developer scoring: FrontrunPro KOL tiers, on-chain track record, 3-layer composite

Phase 2

Agentic research loop: Claude LARP detection, novelty + prior-art analysis

Phase 2

Self-learning model: ridge regression, weekly weight updates

Phase 3

Real-time events: Bitquery indexing + Alchemy webhooks for live claims

Phase 4

Semantic search (Vectorize), trending narratives, RAG chat, swarm sentiment

Phase 4

Creator-identity rework: real developer = fee recipient + project research

Feedback

Found a bug? Have an idea for a new signal? Get in touch — feedback shapes what Seshat tracks next.

Reach out on X ↗