Update Interval of AI Notes

2023-08-31

• In the early days of AI writes research notes daily, each note was updated once a day.

Then, a number of notes could be run in parallel, 22 updates per day.

This was an overload for humans, so we decided to limit the number to 10 per day

• This resulted in indefinite intervals

On the other hand, it is often run directly from the command line with a URL specification

• This is one run now.

There is a feeling that we are not catching the real needs.

Does something that has been executed once really not need a second execution?

• often want to do

Is it really correct to specify the number of sheets and indefinite spacing?

• There's no guarantee that your new loop will be processed tomorrow morning.
• Some want it to be once a day, and some want it to be a little less intense so they can extend the interval more?
  • If you think it's too much, use it.
  • Explicitly switch?
• The use case of "think as much as you can before you go to bed, leave the rest to AI, go to bed, and wake up to find that new developments have occurred" is no longer feasible.
  • In this case, it does not have to be multiple days in a loop
  • One time.
  • Due to the fact that the title cannot be rewritten on import, the style of state management with specific characters in the title will not work.

relevance - AI writes daily research notes: how to select chunks - In nishio, now only those marked 🤖🔁 are eligible for automatic renewal - However, in omoikane and enchi, we used a random target. - In unnamed, "longest page." - Proposal to expand red link - Extending the Red Link with AI - I think this is rather fascinating. - A proposal to find the subject by link or vector search in the first place.

from Diary 2023-09-01 Update Interval of AI Notes Resources on the human side more and more and more and more and more and more metaphor for vegetative growth.

• Let it grow and then prune it

from Diary 2023-09-01 Update Interval of AI Notes 2023-09-01 20:〜 Maybe even a day is too short. Longer term. Grows in a way that engages newly related fragments as they emerge. If so, wouldn't the timing of growth be when the closest fragments appear?

2023/9/2 If it is a long term mode, why not just let it sleep until a closer fragment is discovered? To do so, the proximity of the fragments used must be written in [page memory

• slow-loop

This page is auto-translated from /nishio/AIノートの更新間隔について using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.