Volume 18 [community edition]
The best poker players in NYC; a16z launches Speedrun Alpha + $250k checks; Upscale paint night for founders; Patient-facing healthcare AI is having its moment
📣 Announcements
We’re hosting an exclusive event for the best poker players in NYC
Must be a founder or investor
Nominate yourself or someone you know here 👀
Vol 18 TLDR:
Patient-facing healthcare AI is the next frontier, and how clinical AI actually works: RAG, Bayesian inference, and the trust layer
a16z’s Speedrun Alpha is offering $20K grants and up to $1M in investment
Roles at Sentience, Mercury, Decagon AI, World Labs, and more
📅 Coming Up…
Poker + games
For venture-backed founders + startup engineers
An Evening with Acrylic - Tuesday 3/10
Upscale painting night + cocktails
For venture-backed founders
Poker Night - 3/18
Invite-only. Nominations here.
⚖️ Resources
a16z launched Speedrun Alpha, a fellowship with two tracks
Founder Track
$20K grant, up to $250K investment, automatic final interview @speedrun (up to $1M investment)
Startup Track
Full-time engineering role at an a16z portco
Alliance (nyc based crypto/ai accelerator) just launched ALL17 ($500k checks)
Alliance@allianceThe ALL17 Application Deadline is 1 Month Away3:43 PM · Feb 23, 2026 · 7.46K Views5 Replies · 9 Reposts · 53 Likes
✍🏻 Culture Report: From Billing to Bedside — The Rise of Patient-Facing Healthcare AI
Written by Annie Dong.
The $5 Trillion Problem
In 2024, US healthcare spending soared to over $5T, accounting for 18% of US GDP and outpacing economic growth. Since the launch of ChatGPT in 2022, billions have poured into “healthcare AI.” But nearly all of it has gone to the back office: scribing, prior authorization, revenue cycle management.
The result? Companies like Abridge ($5.3B), OpenEvidence ($12B), and Tennr ($605M), and physicians whose AI usage doubled from 2023 to 2024. The administrative burden is getting lighter. But the patient is still on their own.
Why Patient-Facing AI Is the Next Frontier
The structural problems in American healthcare are human. Cost and access remain prohibitive: rural hospital closures have risen steadily since 2010, and average wait times hit 31 days across the 15 largest US metros. Only 16% of physicians fully exchange and integrate electronic patient information, and the system remains stubbornly reactive. You only see a doctor when something is already wrong.
Five of the top 10 causes of death are strongly associated with preventable, treatable chronic diseases. The system was never designed to catch them early.
The Association of American Medical Colleges projects a deficit of 86,000 physicians by 2036. Patient demand will only grow as the population ages. Burnout is already affecting nearly half of all physicians, and it shows. Nine out of ten US adults struggle to understand their own health information.
Patient-facing AI is positioned to close this gap: democratizing access to medical knowledge, integrating fragmented health records, and giving patients real agency over their own care.
Regulation: From Constraint to Catalyst
The regulatory environment is shifting faster than most realize. In January 2026, Utah became the first state to allow residents to renew medications online through a pilot with Doctronic. In January 2025, the FDA issued its first comprehensive guidance for AI-enabled medical devices across the full product lifecycle. Bill H.R.238 (Healthy Technology Act of 2025) proposed allowing AI to autonomously prescribe FDA-approved drugs at the federal level. Last week, a coalition of health systems launched the first-ever operational standard for AI that communicates directly with patients.
For access to the full article, subscribe here.
⚙️ Under the Hood: How Patient-Facing Clinical AI Actually Works
Written by Priyal Taneja.
Hippocratic AI, Ada Health, ChatGPT Health, Included Health. They all promise the same thing: describe your symptoms, share your records, and the machine guides you toward an answer. Healthcare is a vulnerable domain. What does the black box between your health question and the AI’s clinical response actually look like?
Traditional Clinical Decision Support: The Doctor’s Cheat Sheet
Clinical decision support systems (CDSS) have existed since the 1970s. Early tools like MYCIN used handcrafted if-then rules: if the patient has a fever and a positive blood culture, suggest antibiotics. These systems were rigid, physician-facing, and built on narrow rule sets. They didn’t learn. They looked things up.
Modern CDSS improved on this by layering statistical models over electronic health records, flagging drug interactions and catching abnormal lab values. But they still operated as passive alert engines embedded in clinician workflows, never designed to talk to a patient directly.
What Patient-Facing AI Changes
Patient-facing clinical AI flips the architecture entirely. Instead of surfacing alerts to a physician, these systems conduct a dynamic, conversational clinical encounter with a non-expert user, and do so safely.
At the core is a constrained reasoning loop similar to what powers coding agents: gather information, form a differential, ask clarifying questions, narrow the assessment, recommend a next step. When Ada Health walks you through a symptom check, it isn’t pattern-matching your first answer to a diagnosis. It’s running iterative Bayesian inference. Each answer you provide updates a probability distribution across thousands of conditions, and the system picks the next question that best reduces diagnostic uncertainty.
Reasoning over symptoms is only half the problem. The real engineering challenge is context.
Stitching Together a Fragmented Health Record
The average American’s medical data is scattered across 19 different providers and systems. Patient-facing AI has to ingest and reconcile data from EHRs, pharmacy records, insurance claims, wearables, and lab results, all formatted differently, all using different terminology.
For access to the full article, subscribe here.
🦄 Jobs
Sentience - Fullstack Engineer, AI Engineer, Founding Growth (NYC)
Emanate — Head of GTM
Poka Labs (YC S24) — Founding Fullstack/AI Engineer (SF)
FidoCure — Chief of Staff
Simile — Multiple Roles (SF, NYC)
For access to the full job list (updated weekly), subscribe here.
📷 Photos of the Week
Last week’s game night.
See you next week,
Maggie + Jonas
speedrun loved the collective <3