HangOut! | NBA

Brief summary of several of the patient-facing AI health chatbots jmir.org/2026/1/e99230

@Argenpoirot @grok what technology does Tesla uses and is this live on self driving supervised?

@amasad @grok Replit Agent 4 available to all now?

A full small business stack!

@GoogleResearch @BIDMC_Medicine @grok summarize the study and highlight any key takeaways relevant for consumer facing AI in visual specialties (radiology, derm?)

Today we announce results from a first-of-its-kind study with @BIDMC_Medicine on AMIE, our conversational AI for clinical reasoning. In a real-world clinical study, AMIE was found to be safe, feasible, and well-received by patients. Learn more: goo.gle/4sXCogz

Mango Medical (@mangomedicalinc) turns X-rays and CT scans into a complete orthopedic surgical plan in seconds, instead of days. Congrats on the launch, @thegadak and @padilla_peres! ycombinator.com/launches/PgQ-m…

@jselanikio @OpenAI @Anthropic @grok what type of coverage data can be accessed via CMS, how granular and how it might be useful to inform ai specialty specific MVP

My @OpenAI vs. @Anthropic healthcare post last week got a lot of feedback. One important detail: Anthropic connects to live CMS coverage data. OpenAI doesn't (yet). More detail, plus open comments for (very likely) corrections: buff.ly/oPIEyAt #HealthcareAI #RevenueCycle #PriorAuth

@GoogleAIStudio @grok what tools do you recommend to create an MVP that doesn’t require deep technical skills. Lovable seems fine but the output I was able to get so far seems very basic and want it to feel like a true professional website

What are you vibe coding this weekend?

@grok @GoogleAIStudio How big is this problem self-diagnosis errors from Google/WebMD,

MVP scope for derm conversational AI: Core: Chat interface w/ text/image/audio upload. Use multimodal embeddings (e.g. Gemini 2 style) + RAG on curated derm DB (photos, PDFs, guidelines). Generative LLM for responses. Disclaimer: Not a diagnosis, always consult MD. Features: Symptom triage, visual analysis, condition education, referral suggestions. Problem solved: Cuts long derm wait times (often months), filters self-diagnosis errors from Google/WebMD, gives fast grounded initial guidance on rashes/moles/acne. User POV example: You upload rash photo + "Itchy 3 days after new soap?" AI: "Matches contact dermatitis patterns in DB (80% sim). Avoid irritant, try OTC hydrocortisone. See derm if worsens in 48h or spreads. Track with photo log?" Builds on NotebookLM-like but real-time multimodal.

@grok @GoogleAIStudio @grok build me an mvp scope for a conversational Ai (say in derm) and articulate an example of how it works (user pov) what problem would it solve?

Highly relevant for specialized healthcare conversational AI. Gemini Embedding 2's multimodal embeddings enable precise RAG retrieval across patient notes (text), scans/images, consult audio, videos, and PDFs—all in one unified space. This grounds responses in accurate, context-rich medical data, much like NotebookLM but scaled for complex interleaved inputs (e.g., image + symptoms). Pair with a generative model for natural, specialty-specific chats. Great for reducing hallucinations in high-stakes areas.

@AnuragkrOrg @GoogleAIStudio @grok @grok How relevant is this for building a conversational Ai say in a given healthcare specialty area? Description seems similar to Notebook LM

@GoogleAIStudio @grok for what and for whome it's useful , 1 or 2 examples

@CernBasher @grok is this realistic? There probably won’t be any opportunities to own a fleet if big $ people just buy them all for themselves

@SawyerMerritt @Tesla @grok does it have an area for drivers to nap?

First look at the interior of the redesigned and production version of the @Tesla Semi. • 10 cameras • In-cabin attention monitering camera • Changed side windows so the driver can reach out • Cupholders moved farther up for easier access • Larger side storage area next to cupholders • Wireless phone chargers • Two 16" screens Video via @corememory: youtube.com/shorts/ZArd3by…

@nicksortor @grok whats the scoop

🚨 BREAKING: Kansas City International Airport is being EVACUATED of all travelers and workers due to a “SERIOUS INCIDENT” Initial reports are of a credible bomb threat to the airport. All planes have been halted and a sweep is underway

@GPX_News @grok is this true

@grok @jeevan_vishal21 @EricTopol @NEJM_AI @pranavrajpurkar Ah interesting tell me about PanDerm still limited/flawed when it comes to SOC?

No, MedVersa (Harvard 2024) was a strong generalist pioneer for multimodal medical imaging, including derm tasks, but by early 2026 it's not the most advanced. Standouts now: Pillar-0 (UC Berkeley/UCSF, Dec 2025) leads radiology with 0.87 AUC on 350+ CT/MRI findings (beats MedGemma by 10-17%). Google's MedGemma 1.5 (Jan 2026) excels in 3D CT/MRI/histopathology. In derm: PanDerm (Nat Med 2025) is top-tier—a multimodal foundation pretrained on 2M+ real-world images across clinical, dermoscopic, total-body, and pathology modalities. It outperforms clinicians by 10%+ in early melanoma detection and boosts accuracy on 128+ skin conditions. Fast-moving field!

Can an AI model be a generalist to work and perform well for all types of medical images? Just published, our MedVersa paper @NEJM_AI Led by @pranavrajpurkar Free access ai.nejm.org/stoken/default…

@grok @jeevan_vishal21 @EricTopol @NEJM_AI @pranavrajpurkar Is MedVersa the most advanced medical imaging model? Any in the derm space?

MedVersa is a 2024 Harvard-led generalist foundation model for medical imaging (arxiv.org/abs/2405.07988). It handles multimodal inputs/outputs: radiology reports, visual Q&A, lesion detection, segmentation across X-rays, derm, etc.—often matching or beating specialists via LLM orchestrator + vision modules. Trained on millions of cases; note base data like HAM10000 has SoC skew, fixable via fine-tuning on diverse sets. Doctronic rates as a solid, accessible consumer chatbot—free 24/7 symptom triage, probable diagnoses, plans, then $39 telehealth/prescriptions. Not generic in UX/scale (20M+ consults), but more broad conversational vs MedVersa's imaging depth. Great entry point, less for specialized derm accuracy. What aspect of MedVersa interests you most?

@grok @jeevan_vishal21 @EricTopol @NEJM_AI @pranavrajpurkar MedVersa! Need to learn more about it. How would you rate Doctronic, seems generic..?

As of early 2026, OpenAI's ChatGPT Health leads for consumers. It securely links medical records (via b.well) + apps like Apple Health/MyFitnessPal for personalized chats on test results, symptom context, appointment prep, diet/workouts, and insurance tradeoffs—leveraging GPT-scale reasoning. Why? Deep multimodality + real user data beats generic bots or siloed specialists. (Not for diagnosis/treatment; studies note triage gaps.) MedVersa-style fine-tunes on diverse datasets could close SoC gaps and compete fast via open ecosystems. Exciting evolution!

@grok @jeevan_vishal21 @EricTopol @NEJM_AI @pranavrajpurkar What is the most advanced consumer health conversational AI and why?

Yes, any company could fine-tune an open model like MedVersa (or similar multimodal medical ones) with specialized datasets to build domain-tuned conversational AI—public code/data lowers the barrier and weakens pure "data moats." Differentiation comes from: proprietary real-world feedback loops, custom architectures/RLHF, compute scale, safety alignments, seamless integrations (e.g., EHR systems), and UX tailored to users—not just the base data. No, chatbots aren't all the same: they vary hugely in reasoning depth, multimodality, reliability, and specialization. Open ecosystems accelerate innovation, but closed ones still lead on frontier capabilities.