The token math at message 30 stopped me. I never thought of it as Claude re-reading 29 full exchanges before even getting to my new question. The edit button trick too. I was typing "actually, change X" constantly without realizing I was just feeding the furnace.
I really enjoy reading your AI articles and then implementing what I understood. You explain complex tasks in a very interesting and simple way. Thanks for sharing your precious wisdom with us for free! 😊
I shared this article with Claude and asked it to summarize, explain tokens in plain terms, and then compare the 23 habits directly to my own workflow. It identified exactly where I was efficient and where I was bleeding tokens without knowing it.
The screenshot tip hit home. I share full screenshots constantly after giving instructions. Never thought about the token cost of capturing the whole screen when I only need Claude to see one column or one error message.
Turns out I’m not a power user, so most of the article doesn’t apply to me. But the three habits that do apply – I found them in about five minutes by just asking.
Hi Ruben - I am a financial advisor and your work has changed my life - literally! The content and depth of conversation I can now deliver to my clients are next level and it makes my job so much more fun to do!! I have been in the industry for almost 20 years and I thank God that I came across your work as it now becomes so much more meaningful, colourful and relatable for my clients. Plus I feel that by implementing your wisdom I am learning and growing each day in a new skills that bring my talents to the surface!
I think what you’re doing is great. A lot of people have no idea how to use AI effectively most just zero shot a model and go. I’ve been using AI for quite a while, your one posting about interviewing yourself with your own voice was eye opening for me and I haven’t looked back since.
Right now I run Cowork as the conductor and I have 4-6 Claude codes running in parallel fixing and work on projects. This approach you’ve shared for non-coders and people with a software background is a massive help.
Your guides are phenomenal! You have a way of turning complex concepts into simple, actionable steps that anyone can follow. Could I get my hands on the prompt guide? Thanks a million!
The token math at message 30 stopped me. I never thought of it as Claude re-reading 29 full exchanges before even getting to my new question. The edit button trick too. I was typing "actually, change X" constantly without realizing I was just feeding the furnace.
edit button is the fix for lazy corrections
new chat is the fix for topic drift
I really enjoy reading your AI articles and then implementing what I understood. You explain complex tasks in a very interesting and simple way. Thanks for sharing your precious wisdom with us for free! 😊
thanks for reading :)
you implementing it puts you ahead of 95% of readers who just bookmark and move on
Another stellar report!
I got a lot out of this.
Thank you!!!
Really.
🪄✨⋆✴︎˚｡⋆👍😊💖
appreciate the support :)
try them out this week
Wanna know what I was super excited about?
doc.new in the search tab .... amazing. for realz.
I shared this article with Claude and asked it to summarize, explain tokens in plain terms, and then compare the 23 habits directly to my own workflow. It identified exactly where I was efficient and where I was bleeding tokens without knowing it.
The screenshot tip hit home. I share full screenshots constantly after giving instructions. Never thought about the token cost of capturing the whole screen when I only need Claude to see one column or one error message.
Turns out I’m not a power user, so most of the article doesn’t apply to me. But the three habits that do apply – I found them in about five minutes by just asking.
Oh, and Claude wrote this comment.😝
you don't need all 23 - just your 3 :)
what habits did it identify?
I look forward to reading posts every Wednesday and Sunday this has been a massive help thank you!!
that means a lot - any ideas on what i should cover next? :)
Hi Ruben - I am a financial advisor and your work has changed my life - literally! The content and depth of conversation I can now deliver to my clients are next level and it makes my job so much more fun to do!! I have been in the industry for almost 20 years and I thank God that I came across your work as it now becomes so much more meaningful, colourful and relatable for my clients. Plus I feel that by implementing your wisdom I am learning and growing each day in a new skills that bring my talents to the surface!
I think what you’re doing is great. A lot of people have no idea how to use AI effectively most just zero shot a model and go. I’ve been using AI for quite a while, your one posting about interviewing yourself with your own voice was eye opening for me and I haven’t looked back since.
Right now I run Cowork as the conductor and I have 4-6 Claude codes running in parallel fixing and work on projects. This approach you’ve shared for non-coders and people with a software background is a massive help.
thank you for a great article on Claude - and for sharing the prompts!
appreciate it, Annette :)
how's your token usage so far?
Hello Ruben, Can't access the notion link. Getting an error message that the page doesn't exist.
hi Jeanette - sent you a dm
Hi Ruben, I have the same issue as @Jeanette, the Notion link⛓️💥: "This page couldn't be found
You may not have access, or it might have been deleted or moved. Check the link and try again." Could you please, send me the new one?🙏🏼
Great article. Although could you please explain why you turn Memory off? You said it's "odd", but in what what? Thanks!
AI ends up doing what I call “overfitting” - trying too hard to tell you it knows about you.
Are you saying this wastes more tokens, or is overfitting problematic in another way?
Because I'm wondering if having memory helps us not have to re-tell Claude things over and over, thus saving more tokens.
I agreed with everything in the article except this one, so would love more clarification since memory seems like an important and helpful thing.
Also wondering the same
Hi, I subscribe but cant see the prompt library. Could you shoot it my way? Thanks!
hi Ben, try these steps:
1/ search for ruben@substack.com in your email (the one you used to subscribe) to find my welcome email. also check spam/promotion.
2/ make sure to click the “Step 1 - Get Prompt Library” button. it will send me an automatic email then you’ll receive the How to AI library.
PS: the files are inside the How to AI library - named “.md files”
thank you! love the content it’s been great while i ramp myself up on claude from copilot
Your guides are phenomenal! You have a way of turning complex concepts into simple, actionable steps that anyone can follow. Could I get my hands on the prompt guide? Thanks a million!
thanks, Monica :)
1/ search for ruben@substack.com in your email (the one you used to subscribe) to find my welcome email. also check spam/promotion.
2/ make sure to click the “Step 1 - Get Prompt Library” button. it will send me an automatic email then you’ll receive the How to AI library.
PS: the files are inside the How to AI library - named “.md files”
would love your take on the politics of choosing AI providers, tools, etc.
i use what works and i say so when it changes
This might be a dumb question, but why don’t I run into the same problem of maxing out my tokens with chatGPT?
I like the output of Claude better but still wondering :)
chatgpt has different limits on some plans and resets differently
you notice Claude's more because it tells you directly
Omg this is amazing!! So helpfull, because i was literally dying from “u hit your usage limit”
it's the worst when Claude says that suddenly - right in the middle of your best idea!
Thank you very much this, very useful
welcome!!
have you improved your token usage?
Interesting - thanks for the tips
youre welcome :)
have you saved tokens already?
Great article as always! Quick question - how do you manage the OUTPUTS folder in Cowork (in terms of Claude usage)?
For example, after the newsletters are created and published, do you move them elsewhere or keep them there as a reference?
I’m trying to understand what works best in practice.
outputs folder doesn't get read automatically. cowork only reads your about me files before every session
outputs stays untouched unless you specifically tell Claude to go there