Comet schedules five healthcare ML reading sessions

Deepa Sarin wanted a small daily reading and reflection habit focused on machine learning in healthcare. She connected Comet to her Google account and gave it a task combining article discovery with an explicit calendar artifact.

The request specified five articles, the next five weekdays, a 2pm to 2:15pm window and a reading-session title. Each day was to contain one different article link. Comet searched for relevant material, reasoned through the steps and created the calendar entries.

Sarin checked Google Calendar afterward and reports that all five sessions were present, each with a distinct article attached. This is the completed result: research references were attached to concrete time slots instead of remaining as an unscheduled list in a chat.

The source does not name the five articles, establish an independent quality ranking or show that Sarin later read them. Her report that the task took seconds is not a controlled timing study. Nor does the post explain handling of an existing calendar conflict or a timezone mismatch, so those behaviours are not added to the demonstrated workflow.

The useful pattern is to specify a small corpus, a fixed reading duration and one source per event, then inspect the resulting calendar. The task concerns research scheduling, not clinical advice, healthcare decisions or proven learning gains. The author’s successful calendar check is the human checkpoint actually described.

Prompt
Original prompt

Source: LinkedIn · Deepa Sarin