37 Comments
User's avatar
Dr Sam Illingworth's avatar

Joel, I just want you to know that I thought I did this quite well but then I read your article, extracted it into Notebook LM, worked out the best prompt to put it into Gemini, and it's been a huge help. Thanks so much for showing us the way.

Joel Salinas's avatar

That’s amazing! Goal accomplished ;)

Mariam Vossough's avatar

The relief of not having to re-explain yourself every time you talk to AI is immense. Worth a couple of hours of anyone's time!

Joel Salinas's avatar

Yes!! It saves so much time :)

Chintan Zalani's avatar

Such a ubiquitous setup that will work across AI systems and you laid it down brilliantly. I love the pillars you shared and also how actionable you made them! Thanks for sharing this Joel :)

Joel Salinas's avatar

Glad to hear! Pillars help me grasp concepts well

Mandhir Dua's avatar

For me, takeaway here is that context isn’t a productivity hack, it’s a form of institutional memory. Without it, AI can’t compound learning, only generate isolated responses. Thanks for sharing your views

Joel Salinas's avatar

It really is, thanks for sharing!

Chris Tottman's avatar

Unbelievable time saver. And as you know my thoughts on life in general "all we're doing is burning time " 😳 now I'm burning less time on re-explaining myself

Joel Salinas's avatar

It really is a timesaver! And yes, I think that too haha, wasted time bothers me so much

ToxSec's avatar

“Last 10 minutes: I tested it. Asked a question I’d normally ask. Compared the response to what I usually got. Adjusted preferences slightly.”

slightly is really import. the amount of time i’ve over corrected is too many to count. great stuff here!

Joel Salinas's avatar

Haha same here!!

Alice E's avatar

I asked chatGBT to routinely adopt cross-cultural / cross linguistic concepts in answers - given it does not think in language...

Joel Salinas's avatar

That is so interesting, Alice!! Thank you for sharing!

Joel Salinas's avatar

Never thought of that! And what do you think of how it did?

Alice E's avatar

I only asked it that this evening! I have yet to ask it anything further (I was asking it precisely how it worked re algorithmic bias and wot not, and that came up as an aside)

Joel Salinas's avatar

Excellent! Let me know when you use it some :)

Alice E's avatar

This is what it said to me as an answer to my enquiry:

“For example, if you asked:

“Why do people experience time differently in different settings?”

I might enrich it with:

The Aymara concept of time (past = in front of you; future = behind you).

The Japanese distinction between shōganai (acceptance) and mottainai (regret about waste).

The Greek distinction between chronos (clock time) and kairos (qualitative time).

You don’t have to know the languages — I translate the concept, not the vocabulary.”

Alice E's avatar

I will post here any good results I get...

Jean-Paul Paoli's avatar

Thanks for this share ! You are so right ! Thats exactly the reason why I deactivate everything about automatic memory in ChatGPT and others - I want to be in charge of the context myself and not the LLM. As an example I feel in automatic mode the risk is strong to confus real relevant info vs conversations about potentia hyotohesis.

Joel Salinas's avatar

That risk is real! Hadn’t heard it articulated like this, well said!

Dallas Payne's avatar

Good context is everything! I went through my Claude Skills and all project instructions last week and gave them a good refresh. Made such a difference! Adjusting a Skill eliminated a blindspot I knew Claude gave with certain input. Getting Claude to write its own Skills and project instructions and then tweaking it, is also a really effective way to get it done quickly while still being thorough.

Joel Salinas's avatar

That’s the key, constant iteration :)

Dennis Berry's avatar

Setting it up once turns AI from a generic tool into a teammate who actually gets you

Juan Gonzalez's avatar

You laid out the whole process very clearly.

It's definitely a night and day difference.

Even though I don't let Claude write my content, it provides killer suggestions cause it has the *context* of what I do.

Joel Salinas's avatar

100%! Claude is my cocreator haha

Juan Gonzalez's avatar

Exactly. I now say that I "co-created" something instead of saying that I did it myself. 😆

Michael J. Goldrich's avatar

Context is everything.

You're teaching it who you are, so it stops guessing.

I cover this in my AI readiness work: https://vivander.substack.com/p/the-future-arrived-six-years-early

Joel Salinas's avatar

That’s awesome, thanks for sharing!

John Brewton's avatar

Context changes output more than model size ever will.

Fernando Vago Santana's avatar

Hi, Joel. Which are the files that I should look at in the Premium Hub to provide the context and prepare the AI in the way you suggested? Thanks

Joel Salinas's avatar

Fernando, hello! It will be a combination. For general strategy, check out the resources under Strategic Decision Support: https://leadershipinchange.com/p/premium-member-hub

Also, one of the main benefits is you get to try to implement what's on this post and reach out to me with any questions specific to your case.

Fernando Vago Santana's avatar

Thank you very much

Joel Salinas's avatar

Please send me a private message anytime!

Ruben Hassid's avatar

That’s the whole project setup in a nutshell.

And it’s just two inputs: the prompt before the chat and the files.

Mila Agius's avatar

Critically important article!

I worked hard on it myself and spent quite a lot of time to really understand this approach, but Joel's piece gave me a whole list of tips I hadn't known about before. Thank you so much! Excellent work!