Hacker News new | past | comments | ask | show | jobs | submit login

These sound great in theory but as an app user I have found their implementation can range from deeply irritating to downright broken. Anti-use cases off the top of my head might be (this was from a calendar/reminder app):

- I write myself an appointment on Monday to remind myself to plan a meal for next Friday, so the app silently moves the Monday app to Friday without telling me just because the words "next Friday" appeared in the event title

- I want to quickly note an appointment for a specific time including an address in the title, when the street number is something like 10A - the app aggressively decides (again without warning me) that I want the appointment to be at 10 a.m. instead of the time I originally suggested when creating the event

In the case of that particular app, I just turned off the "natural language" feature globally - shame because it could have been useful if it was less aggressive in its interpretations and silent modifications of my previous choices, but the app developers told me other users are happy with it so there is no reason to fix it.






I think org mode has a great implementation of it. It works when setting a date parameter to, say, "Friday" and the scheduler will pick the closest Friday and turn it into a proper date, so there is no ambiguity later.

To be fair, even a massive LLM is not able to handle this properly

https://chatgpt.com/share/67a8ef08-de44-800a-a0e2-b2523035ee...


Ooh those are good test cases for the calendar event creator with natural language parsing app I'm building :)



Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: