https://t.co/TpJVM3T7Tf built in a week. next 15, app router, postgres, stripe.
the one that cost me two days:
money is stored in micros, not cents. at cents a $5 listing runs 288 ticks in a day and does not move, because 0.031 cents rounds to zero.
it looked like the mechanic was broken. the unit was just too coarse to see it.
every number is public, including mine: https://t.co/DLJ74gCHrY
the burn rates on the rules page are rendered from the same constant the job runs on, so the page and the maths cannot disagree.
$5 gets you on the board.
taking #1 costs whatever the current top has left, plus a cent. that number is on the page and it falls every five minutes, so the crown is cheaper at 4am than it was at midnight.
https://t.co/r3ayvbCoPN
the obvious version is a flat decay: everyone burns 1% an hour.
it doesn't work. a flat rate scales every balance by the same factor, so the ranking never changes. you get a board that drains and still never moves.
the rate has to rise with rank. #1 pays for being #1. that's the whole mechanic.
https://t.co/TpJVM3TFIN
what "nobody holds the top" actually means, with nobody paying anything:
1.25x lead → gone in 14 hours
1.5x → 26 hours
2x → 45 hours
3x → 71 hours
5x → 104 hours
10x → 149 hours
you cannot buy silence here. you can only rent it.
https://t.co/r3ayvbCWFl
@lightsilver323 $5 gets you on the board. :- https://t.co/r3ayvbCoPN
taking #1 costs whatever the current top has left, plus a cent. that number is on the page and it falls every five minutes, so the crown is cheaper at 4am than it was at
midnight.
outbid boards have one bug: someone pays $14,000, takes #1, and the board is dead by
friday.
so i built one where the money burns.
your balance drains every hour. #1 drains fastest, 4%/hr.
a 10x lead is gone in six days without anyone paying a cent.
nobody holds the top. https://t.co/TpJVM3TFIN
Instead of a 3-paragraph "synergistic paradigm shift," the LLM is forced to give you short, clear, non-fluff answers.
Try it out on your next generation.
Tired of LLMs filling every response with "delve," "tapestry," and corporate jargon?
Shoutout to @andrew_n_carr for sharing the ultimate prompt fix to kill AI slop:
Add this to your system prompt: 👉 "Only report to me in ASD-STE100 Simplified Technical English."
Why it works: ASD-STE100 is an aerospace writing standard created for NASA, the US Air Force, and Boeing to keep maintenance manuals error-free. It forces: • Strict limits on word choice and sentence length • Zero metaphors, flowery adjectives, or abstract buzzwords
CLI commands -:
1. pip install preprompt # begin with this *required
2. preprompt-install # one-command setup (API key + hooks) *required
Play around with commands below:
3. preprompt-history # recent prompt events across all sessions
4. preprompt-stats # optimization stats (total, intercepted, avg score)
5. preprompt-memory # learned stack context
6. preprompt-test-classifier # test classifier on sample prompts
7. preprompt-feedback # rate recent optimizations (builds accept/reject stats)
8. preprompt-watch # live feed of interceptions in a second terminal
9. preprompt-clip # optimize clipboard (works on macOS, Windows, Linux)
10. preprompt-optimize "..." # optimize a prompt from the command line