SignalOS
Projects/SignalOS itself

SignalOS itself

The personal exo-cortex this user is building and dogfooding. Continuous-signal ingest → canonical graph → adversarial tribunal → Headway-style insight artifact. Phase 1 = foundation (this PR). Phase 2 = nightly cross-doc discovery + activation spread. Phase 3 = frontend port + DEMO repo deletion.

Vision

A system that reads the world for me, argues both sides on every candidate insight, and surfaces only what would change a decision I'm about to make. Voice is dry, terse, mechanism-grounded.

Strategic Bets

  • Tribunal beats single-critic: adversarial Analyst→Skeptic→Arbiter produces sharper insights than a 7-gate validator
  • Canonical graph beats cosine dedup: explicit alias table + fuzzy match catches what 'Anthropic Inc' vs 'Anthropic' would silently fragment
  • Headway-card beats summary: dense, verb-led thesis + mechanism + named project context > generic 'this looks interesting' digest
  • Per-token activation steering will demote RLHF as the alignment substrate over 12 months

Objectives

  • End-of-Phase-1: signal → Headway insight, deterministic replay works
  • End-of-Phase-2: nightly cross-doc discovery finds one cross-doc insight/week
  • End-of-Phase-3: DEMO repo deleted; frontend ports + insight reels stay

Signals

0

Insights

-

Relevant Signals

View Feed
No signals found. Check ingestion status or system connectivity.