Jokes aside, it’s time to move away from this style of prompting.
For anything that actually runs long hours, multiple context windows, parallel workers, or loops, you need long-horizon-prompting instead.
The new style uses a pseudo-formal task brief:
- Defines exactly what “done” looks like with clear, measurable success conditions
- Explicitly lists what doesn’t count as a win (stops agents from gaming the result)
- Adds an independent fresh-context reviewer that double-checks everything
- Forces the agent to explore different approaches instead of endlessly tweaking the same idea
- Sets strict rules for when to stop and what to actually deliver
I just released the full prompting skill with references here: https://t.co/WUk35nd5NU
Asake should use his hard-earned money to help p00r Nigerians while politicians keep embezzłing public funds?
E get why our forefathers no dey allow women speak for public.