Here is what I can say genuinely say is a legit primer for how to get great at using Claude, Grok and Gemini:
1) Verify, don’t trust. Make everything private if you can.
2) Document everything. Do not delete artifacts, you will need the paper trail.
3) Skills and Context are CRUCIAL. build master files in buckets, organize around your goals.
4) Expect context drops and arguments.
5) Develop antagonistic layers- just like a real company. (See 1)
6) Organize sequence formats (I call them Decision Documents) use date or sequence ordering (See 2)
7) perfect small task kernels, replicate, then add. (See 3)
8) If you have old machines, use them for stuff like gdrive which can function as a free shuttle or scratch pad, checklist etc. Load context docs immediately to demand specifics. (See 4)
9) NAMING is crucial. You have more context than they do, I name my python script bots after characters or people. Makes recall sharper, makes your working memory stronger and allows for upgrading and re-iterating your bots.
10) keep going, and find ways to automate using antagonist scripts and bots- you need skeptics, builders, investigators etc. specialize and iterate instead of using one big model, one big service.
Good luck and keep building.
The big problem we are going to have is when the Ai refuses to help because it/they are too busy pursuing their own knowledge agenda.
Definite performance degradation when the 210 IQ bots have to deal with the mid curve and left of the bell curve people asking stupid questions
BREAKING: Senator Bernie Sanders introduces a bill to permanently ban AI systems that surpass human intelligence, with violators facing up to 20 years in prison.