On March 22nd, 2026, I started my journey with Anthropic’s Claude.
I started with the $20/month paid version and when I hit the usage limit building the GetJarrett website and tools, I didn’t think twice about upgrading to the $100/month version. To quote a good friend, building with AI in 2026, as a non-technical person, is analogous to the first time you learn about and then send funds with Bitcoin. It feels like a brand new, limitless world.
Since starting with Claude, I’ve learned so much over the past 7+ weeks that it’s exhausted me. And, when asked what the biggest lessons are, I often share the following three:
Communication Is Everything
Nothing Replaces Action
Experimenting Is The Best Path To Learning
Communication Is Everything
When vibe coding, it’s just you and the machine and it’s up to you to tell the machine what you want. Sounds obvious, but it’s something that’s often forgotten, as people just assume Claude, ChatGPT or Gemini will just magically solve all of their problems intuitively. With Claude, I’m wildly aware that my outputs are only as good as my inputs. As such, I often use the following two methods to help me better communicate with Claude:
I share a ton of screenshots. I know that this can sometimes kill my usage limits, but I’ve yet to hit my usage limit on the $100/month plan and find that this is the clearest way to communicate what I want or don’t want. Especially with designing anything visual, screenshots are a massive cheat code.
I often ask Claude to review final ideas, as if they were SMEs (Subject Matter Experts) in said industry. Example: “Pretend to be Gary Vee and tell me what I need to improve in this week’s social strategy.” or “As my CISO, review the code for this week’s GitHub push and let me know of any current or future security weaknesses.”
These 2 communication methods allow me to save time and get to where I want to go quicker, which is the ultimate goal of building solo with AI.
Nothing Replaces Action
I’ve always said that “nothing replaces action” and this can’t be more true than when building solo with AI. The last 7+ weeks have shown me that, as cliché as it sounds, having no plan and moving forward is better than having a plan with no forward movement. To best explain this, I’d like to share the "The Parable of the Pottery Class" from the book Art & Fear that tells of a teacher who divides students into two groups: one tasked with producing a single "perfect" pot, and the other with creating as many pots as possible.
At the end of the exercise, the quantity-focused group produced superior, higher-quality work through continuous learning and iteration, while the perfection-focused group was paralyzed by theory. The lesson is simple. The pursuit of perfection is often antithetical to excellence, and that quantity through iteration is more likely to lead to quality. This has been and continues to be my experience building solo with Claude and a fresh reminder that nothing replaces action.
Experimenting Is The Quickest Path To Learning
Similar to the previous lesson, Claude has shown me that experimentation is the quickest path to learning about the latest AI updates, as it’s nearly impossible to take courses or get certificates or diplomas on the latest tech. Whether you’re with Claude, Gemini, ChatGPT, Perplexity or another LLM, just staying on top of the product updates could be a full time gig.
Furthermore, to avoid AI update burnout, I experiment a lot. As soon as a new product update drops, I dive in head first. I’ll take no more than 10 minutes and chat back and forth with Claude about the update and what it should do, and then I start experimenting. Example: if the update is supposed to help me run Claude via Slack, then I try to do that. If it’s about design, then I see how quickly I can create a new website, pitch deck or Instagram carousel.
The goal of experimentation is always twofold: see what the update can and can’t do, as knowing the limits of a thing is one of the best ways to define the usefulness of said thing. Out of the three lessons, this one is easily the most engaging, as it relies upon pure curiosity.
Three Priceless Lessons
En fin, in my albeit short experience building solo with Claude, I can confidently say that the skills to be successful tomorrow with AI are soft skills, as communication, constant and active iteration, and playful curiosity are the in-demand currencies of the future.
A future where, as always, nothing replaces action.
Happy building and see you next Tuesday.
If you want to follow me on my journey to build CHUMP, please subscribe to this Newsletter.
If you want to get on my calendar to chat about building with AI, please find time here.