AI won't suggest a method it didn't think of. and you can't ask for a method you don't know exists. so what do you do?
you tell it to go find those methods first then report to you!
what it suggested vs what existed
when i built my protocols, my claude's answer was a protocol doc - md file - in my vault that it'd read by default
its plan: a protocol doc, a gate (the instructions for when to go read which protocol), and anything that needs reading goes to the vault file
that worked
i mean... was it the best tho?
not really.
it gave me what its 'outdated' dataset had in it
most people don't realize, (most) AI don't search live up-to-date data by default. you gotta make it web search or research. and train it to fall into the habit of searching for latest data before replying
anyway, i randomly found out about session hooks and session start contracts (scripts that load my rules the second a session opens) while just talking to it one day
which would be python scripts. much faster than AI reading a doc or searching thru multiple docs to read what it needs
now my protocols load as a script at the start of every session instead of sitting in md files. SO. MUCH. FASTER.
send it to research before it builds
it only suggests what it already comes pre-loaded with, so before i ask a question or start a build, i have it run live, up to date research on that topic first
my system level instructions also says it needs to run web search and look thru official docs related to a query before it answers
the first time i did it, it wasn't just one high level prompt. i broke it down into very granular things because i wanted to understand how to best run research myself:
crawl Claude's own docs site (official sources)
go through everything in my vault and memories
here's what i need, what to look for, and the kind of upgrades i want
"what else would you add to this research mission?"
#4 is where it adds things i wouldn't have thought to look for
what i do with what it finds
the research comes back with what's available. a list of upgrades. compared to my stack and the processes i run. i don't usually build all of them in one shot.
i go through the suggested terms, improvements it suggests, learn the concepts properly then implement what i feel is necessary
the more control you have on your system the better
Ps. make sure it's reading official sources like release docs, changelogs, API docs, readmes etc.
what you do
have your agent do this before your next build. say:
before we start, research the current official docs for [tool], compare them to how my setup works today, and list anything newer or better i'm not using. then tell me what else you'd add to this research so we can have [xyz or speed, accuracy, token efficiency] upgrades
your homework
pick a workflow you use most. have it run research on latest info on tools that are out there that can upgrade your workflow - from official sources and open source tools available - have your AI learn what it just researched then suggest improvements > build.
PS. when you build, make sure to have your more expensive models route and delegate tasks to your less expensive models. you can use my model-orchestrator from github or run npm install on your terminal:
npm i model-orchestratorread it on the site too:
if you want 1-on-1 help from me... book a call here:
a call to get your AI researching before it builds, with your rules loading the second a session opens
all my calls are here:
hope this helps
SEE YOU ON THE NEXT ONE!
- AUNY 🧡