Skip to content
JJev AtlasField notes
The atlasStart hereClaimsProjectsPatternsIdeasMapEvidenceLibraryFor agents
Search⌘K

J Jev Atlas / Independent research

165 posts · 9 claims · 31 hypotheses

Back to top ↑
All ideas

Build blueprint

Privacy-Preserving Screen Event Filter

A local preflight layer deciding whether a screen change is useful, sensitive, or ignorable.

Authored HypothesisHigh-volume semantic classificationLOW confidenceIndie fit 8/10
Problem
Screen-aware assistants send too much irrelevant or sensitive context upstream.
Why Jev
Every screen event needs fast Boolean and category decisions before expensive processing.
Architecture
On-device OCR/accessibility snapshot → sensitivity/relevance decisions → redact/drop/forward.
Current alternative
Send all frames, sample periodically, or use fragile app allowlists.
Jev advantage
Can reduce both privacy exposure and downstream multimodal cost.
Unknowns
Whether Jev accepts sufficient multimodal-derived state and local-processing requirements.
1–7 day MVP
macOS recorder in shadow mode with a timeline showing kept and rejected events.
Validation experiment
Label 500 screen transitions and measure sensitive false negatives.

Why this confidence: High upside but privacy and input-modality assumptions need testing.

This is an authored hypothesis derived from the research corpus. Nothing here demonstrates product demand, or that Jev performs well on this particular workload. Run the validation experiment before building past the MVP.

Sources

  • officialhttps://docs.typesafe.ai/primitives
  • officialhttps://typesafe.ai/blog/introducing-system-one-models-and-jev
  • socialhttps://x.com/i/web/status/2099928060644749682

Limitations

  • Whether Jev accepts sufficient multimodal-derived state and local-processing requirements.
  • This is a research hypothesis, not evidence of product demand or Jev performance in this workflow.

Supporting research

  • Cascade router PlausibleA cheap decision chooses whether to use rules, a small model, a premium model, a specialist, or a human.
  • HA-Jev ObservedA Home Assistant custom integration exposing Jev evaluations as sensors and automation actions.
  • Cheap decision calls could make tens or hundreds of semantic judgments per event economical. SpeculativeThe product design follows from published pricing and parallel question primitives, not independent production evidence.
  • Parallel decision matrix PlausibleOne state is evaluated against many named questions, replacing repeated prompt/parse cycles.
Record
opportunity:privacy-preserving-screen-event-filter
Canonical
/ideas/privacy-preserving-screen-event-filter
Last verified
2026-09-18