STAT AI: An Emergency Medicine Podcast · Ethan Abbott and DJ Apakama

EP11 Tom Fadial MD,

·44 min·2 clips

No model completed the full SWE-Lancer task set, and the best one earned only about $400,000.

DJ Abhakama opens by placing AI inside the day-to-day reality of emergency medicine. Ethan Abbott joins him in treating the topic as practical, not theoretical. They keep returning to the question of whether new technology will improve the work or just add noise. The guest pushes the conversation toward tool design and the tradeoffs that come with building for clinicians. He argues that suspicious outputs are real and that heavy reliance on AI is a problem. The episode does not treat that warning as a reason to stop. Instead, it asks how to build better safeguards into the workflow. The hosts and guest spend time on the idea that simplicity is not a cosmetic choice. It is a survival strategy when the user is busy and the environment is chaotic. The guest describes making opinionated tools for himself first. That approach keeps the product focused. It also means the result will not suit everyone. He accepts that tradeoff as the price of usefulness. The discussion gets concrete around emergency medicine’s pace and unpredictability. A tool has to fit inside interruptions, uncertainty, and fast decisions. If the interface assumes perfect attention, it fails. If the recommendation layer tries to be everything to everyone, it gets blurry. The guest stresses that expert input helps, but expert-driven design can also become too detailed. The better move is to use expertise as raw material and then simplify aggressively. The group returns to a fracture example and the challenge of deciding how much detail belongs in a recommendation engine. They land on a clear principle: most of the time the tool should just tell the clinician what to do. That creates friction for the edge cases. The episode accepts that friction as part of the job. It closes with a broader reflection on designing for first-time users who will not interact with the product the way the builder imagined. The whole conversation stays anchored in emergency medicine reality rather than AI branding.

As heard by us

Practical AI discussion in emergency medicine, with clear tradeoffs and limits.

This episode keeps its attention on the practical side of AI in emergency medicine. It presents the work as a search for tools that stay simple under pressure, carry enough opinion to be useful, and remain honest about AI's limits.

Read the full review in PlayNext →

Why you'd press play

Want the real-world tradeoffs behind AI tools for emergency medicine?

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from STAT AI: An Emergency Medicine Podcast