OnCall Scribe
A HIPAA-ready patient-visit scribe for small clinics that runs the new Qwen-class open-weight model on local hardware — no patient audio or notes ever leave the building.
family physicians and small-clinic doctors who can't legally send recordings to cloud AI scribe vendors
- On-device transcription and SOAP-note generation using a quantized open-weight model on a clinic Mac Mini or small GPU
- Automatic redaction of patient identifiers before any optional sync to the EHR
- Plug-and-play setup that works with the clinic's existing exam-room microphone
- Offline-first operation so visits continue during internet outages
Qwen 3.8's open-weight 2.4T-parameter release on Hacker News (555 points) signals the tipping point where frontier-grade models can finally run locally — letting clinics ditch cloud scribes and eliminate HIPAA exposure.
Validated, large problem: family doctors spend ~19 hrs/week on admin including 4 hrs of notes (OMA), $600M+ ambient scribe category, Abridge won 2025/2026 Best in KLAS, Canada Health Infoway running largest national AI Scribe Program in primary care globally.AI scribes show promising results in helping family doctors and nurse practitioners ↗Medical AI Tools Landscape 2026 ↗
Market is crowded with HIPAA-compliant cloud offerings (Abridge, Nabla, DeepScribe, Suki, Freed, Nuance DAX/Dragon Copilot, Hippocratic, Notable, Glass Health); Freed/Suki already target solo/small-group at $84-99/mo, and cloud BAAs cover most HIPAA fears — fully-local-only is a narrow niche for the most privacy-paranoid or non-US-regulated clinics.Ambient AI Scribe Comparison Guide 2026 ↗The Most Affordable AI Medical Scribes (2026 Refresh) ↗
WTP is well-validated at $39-$120/mo for self-serve small-practice tiers (Freed ~$99, Suki, others) versus $14K-$30K/yr for human scribes — clear ROI story and easy sell on price.AI Medical Scribe Pricing: What Practices Pay in 2026 ↗How Much Does an AI Medical Scribe Cost? (2026 Guide) ↗
Documentation burden is structural and aging-population-driven, so demand persists; however, the 'no cloud' advantage may erode as BAA-covered cloud scribes mature and regulators accept them — Qwen-class open weights themselves are commoditizing fast.Why family doctors across Canada are turning to AI scribes ↗
Premise is shaky: Qwen 3.8 at 2.4T parameters cannot run on small-clinic local hardware — even quantized, that scale needs datacenter-grade GPU clusters, so the product would have to fall back to much smaller distilled Qwen variants, undermining the 'frontier-grade locally' claim and the HIPAA-by-architecture pitch.Qwen3.8: 2.4T Parameters, Open Weights, No Benchmarks ↗Qwen3.8 Preview: 2.4T Params, Open Weights, Release ↗