AI clinical note writer cites 12/12 real transcript lines, done in 5.4 seconds
MaziyarPanahi · x · 2026-11-13
A developer shared a real doctor-patient consultation test: audio is segmented into per-second bars for doctor and patient voices, and a model writes the clinical note along with a JSON time-to-content map. Crucially, every cited timestamp was checked against the transcript — all 12 citations correspond to real lines, with no hallucinated references. The whole run took 5.4 seconds.
More from Models
- Hybrid Decision Model learns when to reason: 72.5% lower latency at higher accuracy — hbouammar · 2026-11-13
- AI×Bio benchmarks called out as LLM-generated "benchslop" that may be impairing bio-reasoning — anshulkundaje · 2026-11-13
- Tencent Hunyuan internal bench scores its flagship 13.0% on hardest math tier — teortaxesTex · 2026-11-13
- Users report Gemini 3.8 Flash API returning 503 UNAVAILABLE errors for three days straight — Past_Big_2507 · 2026-11-13
- Open OCR costs fell 5x in 18 months: $0.51 to $0.095 per 1,000 pages, benchmarked on 850 pages — vanstriendaniel · 2026-11-13
- Claude outage hits claude.ai, API and Claude Code simultaneously; MCP connectors down — kimmonismus · 2026-11-13