martinthenext

Wow, very similar to what I have been doing but sounds like you had way more consistency. My bot does not remind me to write, it's a neat idea.

I was wondering about adding a trivial Langchain+LLaMa thing to it to chat to my old memories, would be cool to explore this indeed

Yes ofc: https://github.com/martinthenext/daytobase

It is extremely old but probably trivial to upgrade to recent Python. There is also a serverless version [1] and a recently attempted Rust version. I will most likely spend a few hours some weekend writing a Node version soon -- please let me know if you beat me to it :)

[1]: https://github.com/2mol/daytobase

In 2016 I created a telegram bot to make it super easy for me to keep daily diary, and I still use it now:

https://t.me/DaytobaseBot

Similar to other folks in the thread, I sometimes go back to old entries and reflect on how my life has changed. I find that my memory constantly simplifies and idealizes the past, and getting to see how it really felt day to day is very helpful.

Experiments report loss of motivation to do toy tasks. I can imagine people losing motivation to do those, or participate in the study, after meditating. Not sure how that translates to lack of motivation in doing what you actually like/want to be doing.

The nap analogy from the article, I find, is very good. After taking a nap, it might be easier for you to concentrate and do deep work. If you need to do something tedious instead, like respond to admin emails or perform made up tasks for a meditation study, you might perform better pumping some loud music or being under social pressure.

I've been developing something similar, but as a Telegram bot:

https://telegram.me/daytobasebot

https://github.com/martinthenext/daytobase

I found it easier to use telegram because it is available on mobile as well as desktop, so I end up writing more things down.

Pretty great to discover jrnl.sh, validates my idea that this approach to logging things can have wider adoption.

Oh man, we did it before Microsoft!

http://comnsense.io/

https://youtu.be/ALF9GY2K-wc

I was curious and sketched up something similar to this website in about a 100 lines of Python code. It has a CLI interface, have a look if you're interested:

https://gist.github.com/martinthenext/fc989ffa6ec84ee09962

It's pretty interesting how the strategy derived by averaging over multiple intelligent users can be that sub-optimal.