Own Your Attention.

The internet produces more than anyone can read, and the AI systems built to read it for you are owned by a handful of labs and platforms with their own incentives.

Sensorium reads it the way you would — learns what you actually keep, surfaces what's both relevant and new — and runs on your own keys and models, so that you control your feed, not anyone else.

Request access See how it works ↓

Two connected problems

Your attention is under siege

Every feed, newsletter, group chat and podcast is fighting for your attention. Skimming everything is impossible; sampling at random means missing the signal that mattered — and most tools either drown you in more summaries or filter so hard you end up in an echo chamber. Left to chance, the loudest source wins, not the one that actually deserved your time.

Attention is the scarce resource. Sensorium budgets it.

The information layer is being captured

AI answer-engines and algorithmically-controlled feeds are becoming the primary interface between people and reality — and whoever controls that interface gains an unprecedented lever over what populations believe. On current trends, a handful of frontier labs and platforms will own this layer outright, with the capacity to introduce subtle, hard-to-detect steers across the information environment while defenders have no comparable, open tooling.

Independent control is the challenge. Sensorium is yours.

Sensorium Overview

I

Problems addressed

  • Information volume and velocity exceeds individuals' sensemaking capacity — our limited attention is increasingly overwhelmed
  • Provenance / veracity of information hard to ascertain
  • Centralised platform ownership concentrates narrative-setting power
  • AI-driven control over information flows - closed AI systems risk subtle bias and steering
  • Citizens lack tools to verify sources and audit algorithmic information curation
II

Technology solution

  • AI agent web application with browser extension, mobile web app
  • User-controlled, AI-assisted content aggregation, scanning, summarization, curation and sensemaking pipeline
  • Leverages proven "feed scrolling" UX for speed/ease of use
  • Full API for sharing of curated newsfeeds via newsletter / websites / RSS
  • Self-hosted, decentralized deployment architecture
  • Support for diverse open-weight models replacing proprietary API dependencies
  • Future: Fully open-source release, C2PA provenance, auditable logging, Fediverse and AT Protocol integration
III

Outcomes

  • Decentralised deployment removes single points of control
  • Basic version runs on Raspberry Pi 4 upwards (using AI via remote API)
  • User-shareable instances enable community-run sensing nodes
  • Trustless verification through open, auditable provenance standards and logs
  • Faithful AI delegates serving journalists, researchers, and engaged citizens
  • Open epistemic infrastructure as a public good

Sensorium Pipeline

1
Sense every channel, one stream
  • Feeds & publications (RSS/Atom)
  • Social (X, LinkedIn, Bluesky)
  • Email & newsletters from your inbox
  • Podcasts
  • PDFs with OCR
  • Browser clips & phone shares
  • Private messaging (WhatsApp, Telegram, Signal, Discord via Matrix bridges — strictly opt-in per conversation)

Everything reduces to one canonical signal, deduplicated — the same story arriving by three routes lands once.

2
Score relevance × novelty, learned from you

No keyword lists to maintain. Every time you keep something, the interest profile learns. Each signal scores on two axes: relevance (how close to what you decide you care about) and novelty (how different from what you've already seen). High relevance × high novelty is the jackpot; relevant-but-repetitive sinks; novel-but-irrelevant never surfaces.

Every score explains itself — hover any item to see exactly why it ranked.

3
Make sense from articles to claims to conviction
  • Claims, not just summaries — assertions extracted with exact quotes and provenance
  • Stories, scenario quadrants and uncertainties for the bigger picture
  • Hypotheses with evidence ledgers — your scenarios become falsifiable statements whose probabilities move only when evidence arrives

"This scenario went from 20% to 35% this quarter, driven by these seven claims" — a query, not a memory.

4
Act publish what you've distilled
  • AI-assisted summaries, importance calls and citations on every kept item
  • Newsletter-ready digests
  • One-click markdown export
  • Automated posting and sharing to your channels
  • Your own curated RSS feed

Peer-to-peer sharing creates user-controlled epistemic infrastructure.

Own your attention.
Control your reality.

Request access