Hey - It’s Michael.
Had the funniest bus ride the other day - talking to strangers is nice. Enjoy the newsletter!
The Situation
New AI tools pop up every day. Writing tools, research tools, automation tools, sales tools. All promising to 10x your productivity.
If you look closely, most of them are doing the same thing:
They take your input, send it to an LLM, format the output & sell it as a product.
Most AI tools are just wrappers around GPT / LLMs.
The System
The LLM is the engine. But the product is everything built on top:
the interface
the workflow
the constraints
the data it uses
the specific use-case it optimizes for
Essentially, those tools are just:
a well-designed prompt
wrapped in a simple UI
solving a narrow problem
In many cases, you can replicate 80% of the features with a well-written prompt.
(Especially if you remember the issue where I covered prompt engineering.)
In other cases the tool actually saves you hours because it’s been designed exactly for the thing you need.
In my experience however, that’s rarely the case.
In Practice
Instead of asking:
“Which AI tool should I use?”
Ask:
“What exactly do I need? What is this tool doing with an LLM in the background?”
Break it down:
What prompt is it likely using?
What manual work is it removing?
Try to reverse-engineer what the tool is doing - and build a simple version with your own prompt.
You’ll either stop needing the tool or understand why it’s worth paying for.
Both outcomes are useful.
A quote to ponder on:
“The real problem is not whether machines think but whether men do.” — B. F. Skinner
See you next week - Michael