A practising cardiologist has built a patient care application in seven days and secured third place at a hackathon run by artificial intelligence firm Anthropic, a result that is drawing fresh attention to how quickly software development in health care is changing when clinicians begin writing code themselves.
Michał Nedoszytko, an interventional cardiologist based in Brussels, created a post-consultation care system called PostVisit.ai while continuing his clinical duties and travelling between Europe and the United States. His entry was selected from more than 13,000 submissions at the company’s recent developer event, placing it among the top-ranked projects in a competition that drew programmers, researchers, and founders from multiple countries.
The pace of development has become the focus of discussion within medical and technology circles. A decade ago, building even a basic patient follow-up system required teams of engineers, data specialists, and months of planning. In this case, the application was assembled within a week by a single physician using large language model coding assistants and cloud-based development environments.
PostVisit.ai is meant to address a long-standing problem in outpatient care: the drop-off in patient understanding once a consultation ends. Studies in clinical communication have shown that patients often recall less than half of what their doctor explains during a visit, particularly when treatment plans involve medication changes or follow-up tests. This lapse contributes to dosing errors, missed appointments, and delayed recovery.
The application aims to act as a companion for patients after they leave the clinic. It allows users to review notes from their consultation, revisit instructions related to medication or diet, connect health data from wearable devices, and consult published medical literature that relates to their diagnosis. In practical terms, the program attempts to extend the consultation into the hours and days that follow, when most confusion tends to occur.
Nedoszytko is not new to clinical software development. He has previously worked on systems for angiography analysis and hospital department management, and has contributed to academic papers on deep neural networks used in coronary imaging. That background appears to have informed his approach to building a system that fits into existing clinical workflows rather than asking doctors to change them.
Health systems have struggled for years with continuity of care after discharge. A report by the World Health Organisation has linked poor adherence to treatment plans with a large share of avoidable hospital admissions in chronic disease management. In cardiology, where missed doses of anticoagulants or statins can lead to severe outcomes, the cost of misunderstanding instructions can be high.
Technology companies have attempted to address this gap through patient portals and reminder systems. However, uptake has been uneven, often because these systems rely on manual input or present information in a way that patients find difficult to interpret. The idea behind PostVisit.ai is that conversational interfaces can interpret a treatment plan in plain language and respond to patient questions in real time.
Cardiologist wins 3rd place at Anthropic’s hackathon. Out of 13,000 applications. Built in 7 days by Michał Nedoszytko MD. Coded day and night – in the hospital, in the cloud, while flying from Brussels to San Francisco.
A few years ago, it would have been impossible for a doctor… pic.twitter.com/nNtf9mnmfH— Michał Podlewski (@trajektoriePL) February 20, 2026
The hackathon result has also highlighted a broader trend: professionals outside the software industry are increasingly building applications that once required formal engineering teams. Recent surveys of medical staff in Europe and North America suggest that doctors are experimenting with code-writing assistants to automate administrative tasks such as discharge summaries or insurance documentation.
Anthropic’s competition encouraged participants to create applications that use its large language models to manage extended text inputs, such as medical histories or clinical guidelines. The ability to process long patient records has been a major limitation for earlier systems, which often struggled to maintain coherence across thousands of words of clinical data.
In demonstrations shared online, PostVisit.ai appears to allow patients to upload discharge summaries, laboratory results, or imaging reports and receive explanations in everyday terms. It can also link to data from consumer health devices, though questions remain about how such information should be interpreted in a medical setting.
Privacy remains a concern. Systems that handle patient records must comply with strict data protection rules in Europe and elsewhere, including provisions that limit the sharing of identifiable health information. Nedoszytko has stated that encryption and controlled access protocols were built into the application, but an independent review would be required before any hospital could integrate it into routine practice.
Accuracy is another issue that experts continue to raise. Large language models can produce convincing responses that are not always clinically correct, particularly when asked about rare conditions or unusual drug interactions. While PostVisit.ai attempts to reference published literature in its replies, such features would need validation in controlled trials before clinical use.
