“I see you doing all this AI stuff. I want to start. I just don’t know where.” Five of my friends sent me almost word-for-word that message in the last two months. So I’m putting it here, in one place.
Two questions before the toolkit
The reason “where do I start” is so hard is the choice space. Hundreds of tools, every thread on X claiming a different one is the only one. Exhausting before you’ve opened anything. Before any tool, answer the first question.
1/ What’s your angle to start leveraging AI more?
The work angle. Corporate job, you want to automate workflows or look sharper at your next review. Time horizon: days. Constraint: your company’s approved-tools list. Inside that fence, quick wins are everywhere.
The pet project angle. Something small for yourself - a personal site, a tiny automation, a side experiment. Time horizon: weeks. Total freedom on tools. The trap: infinite choice eats more weekends than building does.
The entrepreneur angle. Ship a real product end-to-end, get real users on it. Time horizon: months. My honest advice: go slowly at the start. The slower you move now, the further you’ll go in six months.
Pick the one actually true for you. Then there’s the second question.
2/ Foundations or wrappers?
Foundations
They are the raw models and their first-party apps. OpenAI ecosystem (GPT, Atlas, Codex) as an example. Covers almost any task you can describe in words. Lots of freedom. Almost nothing back if you don’t know how to ask. Once learned, you own everything you make in it.
Wrappers
Specialised services on top of those same foundations:
Framer or Lovable for a website.
Rork for a mobile app.
Gamma for a slide deck.
Higgsfield for visuals.
etc.
Same underlying AI, dressed up by engineers for one job. You learn a wrapper in an afternoon. A foundation takes weeks.
The tradeoff
The way I think about this trade of speed vs depth is clothes.
A wrapper is the T-shirt that fits you perfectly today.
A foundation is a shirt two sizes too big. Weird at first. But the kid in the perfect T-shirt outgrows it. The kid in the oversized one keeps wearing the same outfit, and one day it just fits.
You’re not picking what fits you today. You’re picking what’ll fit the version of you six months from now (if you intend to play long-term)
Which side you start on falls straight out of Question 1.
Work goes wrapper nine times out of ten – speed and risk are the constraints. A dashboard from a corporate Replit to impress colleagues tomorrow.
Pet project goes wrapper first for the quick win, climbs to a foundation when the ceiling shows up.
Entrepreneur eventually needs foundations - the earlier you start that curve, the more it compounds.
I came in from the learning angle, eyeing the entrepreneur side. I wanted to understand how this stuff works and own the technological layer instead of renting it. Free to pick anything, I leaned hard toward foundations.
Claude became my pick. Two reasons:
A fresh start. ChatGPT had been casual for two years and felt stale.
A growth trajectory. I’d heard about this thing called Claude Code, then a few months old, and figured it’d give me room to grow into.
I started the way I described in Issue 1 – chatting with Claude about what I wanted to build. The first code snippets I pasted into a real file, the moment it actually worked… insane.
Months in, my stack is still 90% Claude. Mostly Claude Code. I’m glad I went slow instead of grabbing a Lovable or a Rork or a Replit. Slow let me understand more, own what I do, and adjust things I’d have missed otherwise.
An Idea For This Week
This time the idea is reflection with an action chaser. I’d sit with two questions and write the answers down somewhere physical:
What am I actually trying to solve with AI right now? (Not “be more productive” - something specific I’ve been avoiding this month.)
Am I the wrapper kind - someone who wants a polished result this weekend? Or the foundation kind - willing to spend a few weeks on the awkward shape of a tool, in exchange for owning what I build with it later?
Then pick one tool that fits both, block 30 minutes, open it. Pinning the questions down is half the work. Acting within seven days is the other half.
The wrong tool you actually use beats the perfect tool you keep researching.
This Week I Shipped
A new user came from Claude recommendation
This week I caught a moment I’d been waiting months for. A new Soka user signed up, uploaded a test, and left a five-star review - and they hadn’t come from my marketing. Claude had recommended one of my blog articles. They read a few more, then came in.
SEO is starting to pay off. It really does feel like getting clients while you sleep. Doubling down on organic across all projects.
Operator or foundation - which side did you land on, and what made you pick it? Reply on this e-mail, I read everything