Independent testing will determine whether Premsys delivers practically useful fully on-device Gemma inference, private RAG, document workflows, web search, and native iOS integrations at acceptable quality and latency.
state: expiredheat: lowuncertainty: highknownscott: mediumlocal-inference on-device-agents mobile-ragPremsys
What is this?
Premsys is an iOS AI assistant that claims to run Gemma models fully on-device, without accounts or datacenters. Its official app page says it can build a local RAG index for private document search and access Calendar, Health, Reminders, Weather, and web search through permission-controlled integrations. The supplied evidence comes primarily from Premsys itself and does not establish independent testing, real-world output quality, latency, or the practical reliability of these workflows.
Why it matters to Scott
The validation claim is already held in Scott’s Evaluation-Driven Development page: product assertions about local inference, RAG quality, and latency require measured use. Premsys is nevertheless relevant as a testable iOS implementation touching his hardware-aware local inference and Apple Health evidence-retrieval work, while the radar already tracks closely parallel validation cases for Orvena’s iPhone agent harness and Minna’s local RAG.
ip:concept.evaluation-driven-developmentdev:concept.hardware-aware-local-inferencedev:project.apple-healthradar:orvena-4b-on-device-agent-harnessradar:minna-local-rag-validationradar:concept.mobile-agents
queries asked of Scott's wikis
- on-device inference quality and latency thresholds
- private mobile RAG architecture
- local-first AI and data sovereignty
- agent tool permissions on iOS
- on-device agents versus cloud agents
- mobile document indexing and retrieval
Measured heat
no measured readings yet — the hourly heat pass fills this in
How the heat travelled
no chain yet — the hourly chain pass fills this in
Evidence (2) — ⭐ canonical anchor
Interpretation history
2026-08-24T06:22:17Z
The launch episode has faded without independent testing, measured performance, or implementation follow-through. Premsys remains a testable vendor artifact, but it provides no new evidence for practically useful on-device agents or mobile RAG.
2026-08-22T05:28:16Z
No independent testing has appeared after the initial release window, so Premsys remains a testable but wholly vendor-asserted implementation rather than evidence of practically useful on-device agents or mobile RAG.
2026-08-20T05:27:55Z
No independent testing or implementation evidence has arrived; the case remains a released but unvalidated on-device iOS assistant, with only repetitive engagement reobservation.
2026-08-20T05:27:10Z
grounded: known/medium — The validation claim is already held in Scott’s Evaluation-Driven Development page: product assertions about local inference, RAG quality, and latency require m
2026-08-20T05:24:00Z
origin walked (codex/luna, conf 0.94): anchor reddit.post.1vt9epk -> echo.other.4aabc6f1c7 by Premsys Inc.
2026-08-20T05:22:53Z
case created — The released iOS app is a concrete, independently testable local-agent artifact distinct from existing on-device harness and RAG cases.
Decision trace
- 08-24 16:22expireThe launch episode has faded without independent testing, measured performance, or implementation follow-through. Premsys remains a testable vendor artifact, but it provides no new evidence for practi
- 08-24 16:22alert_silentThe staleness trigger adds no consequential evidence; any future independent benchmark or workflow test can reopen the subject as a new delta.
- 08-24 16:22alert_routeThe staleness trigger adds no consequential evidence; any future independent benchmark or workflow test can reopen the subject as a new delta.
- 08-22 15:28repriceNo independent testing has appeared after the initial release window, so Premsys remains a testable but wholly vendor-asserted implementation rather than evidence of practically useful on-device agent
- 08-22 15:28alert_silentThe staleness trigger adds no consequential evidence; wait for measured latency, quality, privacy verification, or reliable workflow testing before surfacing it.
- 08-22 15:28alert_routeThe staleness trigger adds no consequential evidence; wait for measured latency, quality, privacy verification, or reliable workflow testing before surfacing it.
- 08-20 15:27repriceNo independent testing or implementation evidence has arrived; the case remains a released but unvalidated on-device iOS assistant, with only repetitive engagement reobservation.
- 08-20 15:27alert_silentThe only change is a negligible Reddit vote-ratio movement, which adds no evidence about local execution, privacy, RAG quality, latency, or integration reliability and can wait for routine review.
- 08-20 15:27alert_routeThe only change is a negligible Reddit vote-ratio movement, which adds no evidence about local execution, privacy, RAG quality, latency, or integration reliability and can wait for routine review.
- 08-20 15:27alert_silentThe App Store release is established and creates a testable iOS implementation of fully local Gemma inference, private RAG, Health data retrieval, and native tool integrations, but those capability, p
- 08-20 15:27surface_candidateThe App Store release is established and creates a testable iOS implementation of fully local Gemma inference, private RAG, Health data retrieval, and native tool integrations, but those capability, p
- 08-20 15:27alert_routeThe App Store release is established and creates a testable iOS implementation of fully local Gemma inference, private RAG, Health data retrieval, and native tool integrations, but those capability, p
- 08-20 15:27groundThe validation claim is already held in Scott’s Evaluation-Driven Development page: product assertions about local inference, RAG quality, and latency require measured use. Premsys is nevertheless rel
- 08-20 15:24promote_anchororigin walk conf 0.94
- 08-20 15:22createThe released iOS app is a concrete, independently testable local-agent artifact distinct from existing on-device harness and RAG cases.