I have one controversial comment. You previously said we could use ChatGPT or Gemini as a teacher (study mode), which means we don't know the topic. Now, in the last few newsletters, you've been advising for the opposite: use AI only in topics that we master.
But let's say I I'm an expert in the medical field and I want to run my consultancy company. Does that mean that I cannot use any AI tool to help me with the business plan because I wouldn't know how to differentiate correct from incorrect information in the business field?
There's a difference between learning a topic and executing on a topic.
Study Mode? Perfect for learning. You want to understand business fundamentals. You want to know what a P&L looks like. You want to grasp pricing strategies. AI can teach you that. It's a great teacher for concepts.
But execution is different.
When you actually write your business plan, you need judgment. You need to know if the AI output makes sense for your specific situation. You need to spot when it's giving you generic advice that doesn't fit your market.
So here's the answer to your example:
Use AI to learn the concepts. What goes into a business plan. What questions to ask. What a good financial projection looks like. Let it teach you the fundamentals.
Then use AI to draft the plan, under your supervision, with your judgment on what feels right for your specific situation. You might not be a business expert, but you know your market. You know your clients. You know what services you can deliver.
The mirror vs teacher thing is don't let AI make decisions for you in areas where you have no way to evaluate if it's right.
Hi Ruben, test this approach for your own business but now as a tax agent. You will soon find out you will need to get registered in many countries where your readers are to remit local VAT out of your subscription. Now try to register e.g. in Poland without spending a fortune on fees. It will drive you crazy. 🤪
Wow! I needed this. I am a reader, bookmark for later. Attempted many and am mediocre at best. Ha! For work I play a customer success manager. Recently got certified as a health and wellness coach. I need to stick to one and just do it!!! Mess up. Fail. It’s ok. Ruben said so. Many thanks for this letter. Will share.
You already know the problem. That's half the battle. Most people bookmark forever and never admit it's decision paralysis. Pick one tool. Use it for your coaching clients this week. Not for work. Not for both. Just coaching.
Report back here!
That's actually why I have a private Slack (with paid members) so I can be an accountability buddy for you guys.
This one stung in a good way. Reading about AI feels like progress, but it’s mostly just staying comfortable. I can feel busy and informed, but nothing really actually changes or there is no action to it. The overwhelm isn’t the speed of AI, it’s the gap between reading and doing. Also, “show, don’t ask” is the line I’m taking with me!
Not sure if this is the right place to ask this. I'm a paid subscriber for the newsletter & I can't find the email that has the links to the resources Ruben shared (& I think access to the private slack). How would I go about having that re-sent. Thanks!
I think if people feel behind in AI and are reading your newsletter than they are either an innovator or early adopter. My mom and those in my neighborhood don’t feel behind in AI. They are concerned about AI, and most have tried it. Many have not downloaded chat gpt yet.
It's also for people who feel behind but shouldn't be. Who have every advantage but aren't using it.
I've seen it hundreds of times. People who know about Claude, know about Projects, know about memory, know about Custom Instructions. They've read every announcement. They can explain the difference between Opus and Sonnet.
They've never actually used any of it - informed non-doers.
I see it constantly. Someone wants to use AI to draft a proposal. So they schedule a call to align on the approach. Then another call to review the plan. Then a third call to get sign-off.
Meanwhile, the actual draft - twenty minutes with Claude.
Now you can show the thing before anyone asks for the plan.
Well done and thank you, Ruben, for this insightful, wise, and even moving article. This step back is beneficial, I agree. Furthermore, your recommendation to test things and form your own opinion seems to me to be the cornerstone of learning about AI. There are too many announcements, too many products, too many possible use cases. You have to start with your own needs and try to find your own solution, however you can. Even though AI is very powerful and relatively easy to access, it can't shorten the time needed for personal experience. And you have to set aside that time in your schedule to progress and build a real understanding of AI. Okay, I'm going back to experimenting...! :)
Challenge… if everybody follows your advice, publish first then see what happens, will the process globally collapse because of overload of new stuff to digest. I am aware this might just be my anxiety. Great thought piece Rubin.
Fair question. But no, the process won't collapse. I'm not saying publish everything to the world. I'm saying use AI to test ideas fast, then show up to your meeting with the work done instead of asking for permission first. The filter is still your taste. AI gives you speed. You decide what ships publicly. If everyone had good taste and tested more versions before sharing, we'd actually have less noise, not more. The anxiety you're feeling is normal. Just start small with one project and see what happens.
Thank you for responding. Any work that is done can be tested against the algorithms of compressed data in the latent space of AI. If this is what you are referring to we could be on the same page and maybe should speak privately to compare notes
Let me ask you a question. Do you have any training material you have used about processes around Machine Learning data compression and/or game theory role in inference? What do you use to determine if something is worth publishing?
Really appreciate that. Seriously means a lot. I'm just someone who got obsessed with this stuff three years ago and kept going. The only difference between me and you is I started earlier and documented everything along the way.
And PS, if you love a good upbeat soundtrack, this couple are the most tuned-in deep house DJs (fused with disco, funk, R&B...) I've ever found. This is a great set, but their others are usually in front of dramatic vistas all over the place. Remarkable stuff. Enjoy. https://www.youtube.com/watch?v=CAg8SKjVDCY&list=RDgVF00MxZzS8&index=2
love that you read comments, Ruben—they must be a flood, and will only get deeper. As to our arcs, I also got in 3 years ago, but clearly could not dive as deeply as you. But I do have quite an archive documenting and prioritizing everything I practice and think are important—most of it from you! Be well, sir. You're a marvel.
Loved loved this article, I felt as if you were talking about me — except I’m lucky, I’m retired. So no worries about my job. I can play like a kid ALL DAY LONG ( so long as I keep the wife happy)
You're in the best position possible. No politics, no boss watching over your shoulder, just pure experimentation. Most people can't afford to test everything because they're worried about looking stupid at work. You can break things and nobody cares. That's the dream setup. Just make sure you're actually shipping something, not just collecting bookmarks. Test one thing, use it until it's boring, then move to the next. And if you figure out something that works, write it down.
The best teachers are people who just learned something yesterday. So you're onto something!!
This was quite a funny newsletter!
I have one controversial comment. You previously said we could use ChatGPT or Gemini as a teacher (study mode), which means we don't know the topic. Now, in the last few newsletters, you've been advising for the opposite: use AI only in topics that we master.
But let's say I I'm an expert in the medical field and I want to run my consultancy company. Does that mean that I cannot use any AI tool to help me with the business plan because I wouldn't know how to differentiate correct from incorrect information in the business field?
You caught me.
There's a difference between learning a topic and executing on a topic.
Study Mode? Perfect for learning. You want to understand business fundamentals. You want to know what a P&L looks like. You want to grasp pricing strategies. AI can teach you that. It's a great teacher for concepts.
But execution is different.
When you actually write your business plan, you need judgment. You need to know if the AI output makes sense for your specific situation. You need to spot when it's giving you generic advice that doesn't fit your market.
So here's the answer to your example:
Use AI to learn the concepts. What goes into a business plan. What questions to ask. What a good financial projection looks like. Let it teach you the fundamentals.
Then use AI to draft the plan, under your supervision, with your judgment on what feels right for your specific situation. You might not be a business expert, but you know your market. You know your clients. You know what services you can deliver.
The mirror vs teacher thing is don't let AI make decisions for you in areas where you have no way to evaluate if it's right.
That makes complete sense. Thank you so much for answering my question!
Hi Ruben, test this approach for your own business but now as a tax agent. You will soon find out you will need to get registered in many countries where your readers are to remit local VAT out of your subscription. Now try to register e.g. in Poland without spending a fortune on fees. It will drive you crazy. 🤪
Wow! I needed this. I am a reader, bookmark for later. Attempted many and am mediocre at best. Ha! For work I play a customer success manager. Recently got certified as a health and wellness coach. I need to stick to one and just do it!!! Mess up. Fail. It’s ok. Ruben said so. Many thanks for this letter. Will share.
You already know the problem. That's half the battle. Most people bookmark forever and never admit it's decision paralysis. Pick one tool. Use it for your coaching clients this week. Not for work. Not for both. Just coaching.
Report back here!
That's actually why I have a private Slack (with paid members) so I can be an accountability buddy for you guys.
“I just read myself again, and I’m so dramatic, I really need to chill.” Ahahah yes but it’s part of the charm
I really am like this though ahaha and you know it
This hit hard. I've been watching AI news from X. And it’s really overwhelming. All of it. Feeling behind every single day.
Never realized I was living on the wrong clock the whole time.
I'm not slow and certainly not behind.
One ugly attempt is how I’ll start my week starting today.
That guilt was real. But it was misplaced. You live on adoption speed.
When did your parents actually start using smartphones daily? 2012. Maybe 2014 :)
Go make something ugly this week. Then make something slightly less ugly next week.
This one stung in a good way. Reading about AI feels like progress, but it’s mostly just staying comfortable. I can feel busy and informed, but nothing really actually changes or there is no action to it. The overwhelm isn’t the speed of AI, it’s the gap between reading and doing. Also, “show, don’t ask” is the line I’m taking with me!
You're comfortable enough that nothing feels broken. But nothing feels great either.
You're spectating. And spectators don't get ahead. Doers do.
One ugly attempt. One scrappy draft. Do it wrong.
Have fun doing it wrong :)
Not sure if this is the right place to ask this. I'm a paid subscriber for the newsletter & I can't find the email that has the links to the resources Ruben shared (& I think access to the private slack). How would I go about having that re-sent. Thanks!
Sure - I sent you a dm!!
Okay, do what Ruben said..."Tom, don't bookmark, DO"...repeat..."Don't bookmark, DO"....repeat.....Okay, okay, I got it!!! :)
Thanks, Ruben, I needed to read this!
Let’s keep being the doer!!
Any preference for future newsletters? :)
I think if people feel behind in AI and are reading your newsletter than they are either an innovator or early adopter. My mom and those in my neighborhood don’t feel behind in AI. They are concerned about AI, and most have tried it. Many have not downloaded chat gpt yet.
It's also for people who feel behind but shouldn't be. Who have every advantage but aren't using it.
I've seen it hundreds of times. People who know about Claude, know about Projects, know about memory, know about Custom Instructions. They've read every announcement. They can explain the difference between Opus and Sonnet.
They've never actually used any of it - informed non-doers.
The permission loop is the real killer.
People spend more time planning the AI workflow than just trying it.
Stop asking if you should use AI for something. Just use it. Show the result. Then decide if it's worth keeping.
I see it constantly. Someone wants to use AI to draft a proposal. So they schedule a call to align on the approach. Then another call to review the plan. Then a third call to get sign-off.
Meanwhile, the actual draft - twenty minutes with Claude.
Now you can show the thing before anyone asks for the plan.
Ruben, love your take because you name the real pain: AI headlines sprint, my actual Tuesday crawls.
The three clocks framing is the keeper. Tech speed, app speed, adoption speed. That alone lowers the panic.
The traps also sting because they’re true. Permission loop, news addiction, tool carousel, old reflexes. I’ve lived in at least two of these lately.
One tweak: velocity isn’t the whole game. Velocity plus a tight feedback loop is. Ship something small, measure one thing, adjust one thing, repeat.
Also, “AI is the mirror, not the teacher” is great. I’d add: it becomes a teacher when you force it into critique, iteration, and constraints.
Net: this makes people stop consuming and start doing. Rare.
That's the actual game. I implied it with "show, don't ask" and "play like a kid". But you made it explicit. Better.
Yes, AI becomes a teacher when you force it. Most people don't do that. They ask AI to write something. AI writes something. They copy-paste. Done.
No feedback. No iteration. No constraints. Just output.
That's using AI like a vending machine. Insert prompt, receive mediocre result.
Iteration is so key and too many don’t even think of giving an AI output a second pass or more.
Crazy to think how fast the technology is advancing. And the funny part is, it’s never going to be as slow as it is now.🩷🦩
We’re living on adoption speed. Where change happens slowly, awkwardly, and with a lot of friction.
One tool. One hour. One task. One ugly attempt without expectations.
Well done and thank you, Ruben, for this insightful, wise, and even moving article. This step back is beneficial, I agree. Furthermore, your recommendation to test things and form your own opinion seems to me to be the cornerstone of learning about AI. There are too many announcements, too many products, too many possible use cases. You have to start with your own needs and try to find your own solution, however you can. Even though AI is very powerful and relatively easy to access, it can't shorten the time needed for personal experience. And you have to set aside that time in your schedule to progress and build a real understanding of AI. Okay, I'm going back to experimenting...! :)
Doing is the only way. Keep being a doer!!
Feeling behind is normal when you watch breakthroughs but live on adoption speed.
it’s sometimes hard to keep up
Challenge… if everybody follows your advice, publish first then see what happens, will the process globally collapse because of overload of new stuff to digest. I am aware this might just be my anxiety. Great thought piece Rubin.
Fair question. But no, the process won't collapse. I'm not saying publish everything to the world. I'm saying use AI to test ideas fast, then show up to your meeting with the work done instead of asking for permission first. The filter is still your taste. AI gives you speed. You decide what ships publicly. If everyone had good taste and tested more versions before sharing, we'd actually have less noise, not more. The anxiety you're feeling is normal. Just start small with one project and see what happens.
Thank you for responding. Any work that is done can be tested against the algorithms of compressed data in the latent space of AI. If this is what you are referring to we could be on the same page and maybe should speak privately to compare notes
I prefer comparing notes publicly so 200,000+ people can benefit from it.
i think we are talking about two different things
Let me ask you a question. Do you have any training material you have used about processes around Machine Learning data compression and/or game theory role in inference? What do you use to determine if something is worth publishing?
SIMPLY BRILLIANT. @Ruben, you surprise me and impress me more every day. I have great confidence you are a birthing star.
Really appreciate that. Seriously means a lot. I'm just someone who got obsessed with this stuff three years ago and kept going. The only difference between me and you is I started earlier and documented everything along the way.
And PS, if you love a good upbeat soundtrack, this couple are the most tuned-in deep house DJs (fused with disco, funk, R&B...) I've ever found. This is a great set, but their others are usually in front of dramatic vistas all over the place. Remarkable stuff. Enjoy. https://www.youtube.com/watch?v=CAg8SKjVDCY&list=RDgVF00MxZzS8&index=2
oh super cool playlist too!
love that you read comments, Ruben—they must be a flood, and will only get deeper. As to our arcs, I also got in 3 years ago, but clearly could not dive as deeply as you. But I do have quite an archive documenting and prioritizing everything I practice and think are important—most of it from you! Be well, sir. You're a marvel.
thanks Scott!! i do my best to comment back.
your post today is a very detailed challenge that I cannot delay acting on. It's like your writing is a perfect prompt to my internal AI!
Loved loved this article, I felt as if you were talking about me — except I’m lucky, I’m retired. So no worries about my job. I can play like a kid ALL DAY LONG ( so long as I keep the wife happy)
You're in the best position possible. No politics, no boss watching over your shoulder, just pure experimentation. Most people can't afford to test everything because they're worried about looking stupid at work. You can break things and nobody cares. That's the dream setup. Just make sure you're actually shipping something, not just collecting bookmarks. Test one thing, use it until it's boring, then move to the next. And if you figure out something that works, write it down.
The best teachers are people who just learned something yesterday. So you're onto something!!