Systems | Sitanshu Sinha

The systems, at system level.What each system does, how its loop closes, and what proves it worked. No client names, no logos: the systems stand on their own.Flagship: FactLens, a working fact-check pipeline, built on one Mac mini and run on real viral posts.Sibling: Nudger, a media-diet judge that classifies every post I scroll and nudges me when the mix tips, built on the same Mac mini.Build infrastructureThe scaffolding every other system runs on: how a build is planned, dispatched, and verified.

An agentic operating system: A single backend that coordinates specialized capability modules behind one API.

A constitution for AI-built software: A written spec before code, a mechanical exit gate, and a second reviewer who cannot edit.

Intelligence systemsSystems that read, reason, and simulate: narrative, audience, and operations.

A narrative-intelligence engine: Watches many public surfaces at once, classifies brand-relative, and abstains when the evidence is thin.

A synthetic-audience simulator: A weighted panel of synthetic personas estimates reaction, audited against its own bias before it is trusted.

A creator-operations platform: Chat-first coordination between a creator and a production agency, with one hard approval gate no message skips.

Personal systemsSystems built for one person's real problems: money, research, and memory.

A forensic personal-finance ledger: Ingests statements into a deduplicated, ownership-verified ledger, and never assumes a purchase is yours.

A research second brain: A topic in, a grounded report and a chat interface out, filed as plain markdown in a personal vault.

A face-first event gallery: A folder of event media and a folder of reference headshots in, a browsable face-organised gallery out.

My Real Inflation: set your basket, see your rate.