Rendered at 12:21:21 GMT+0000 (Coordinated Universal Time) with Vercel.
hemc4 9 hours ago [-]
I can't write any prompt without typos, never thought it will increase my LLM cost. Time to fix this habit.
I am gonna keep my "please" and "thanks" though. Trying to save few tokens can lead to changed habit in real life.
rgovostes 9 hours ago [-]
I appreciate that the post was short and to the point, but it is, like so much content submitted to HN now, heavily filtered through an LLM's voice, if not completely written by one.
> Writing things down generally helps me build my own clarity and I hope to get the same out of this particular write up. Hopefully I achieve this clarity sooner rather than later.
Analemma_ 11 hours ago [-]
Are people really trying to cut the token counts of their manually-typed prompts? Obviously it's good to do what you can to trim skills, documentation that gets ingested, or anything which is in an automated pipeline, but trying to police your typing like this in a live chat session just seems like making yourself crazy to save a fraction of a cent.
542458 10 hours ago [-]
Agreed. Saving maybe a dozen tokens is meaningless when a task can easily chew through ten thousand times that many. A single misdescribed task will use more tokens than all my spelling mistakes all year.
I am gonna keep my "please" and "thanks" though. Trying to save few tokens can lead to changed habit in real life.
Here is how the author used to write: https://pankajpipada.com/posts/2022-07-09-voraciousness/
> Writing things down generally helps me build my own clarity and I hope to get the same out of this particular write up. Hopefully I achieve this clarity sooner rather than later.