Back again with another W as i look forward to my 21st birthday this Saturday. Thanks to Nate's YT, Claude Code and building meaningful solutions for businesses. Go to my previous W you will see the deposit. This is final payment Hope this inspires others and never give up AI is really making it possible for anyone creative, hard working and willing to stand the pain and solve problems to eventually win, super grateful as well.

I built my first ever skill. What I built: a skill that drafts my invoices and letters of agreement. This is admin I repeat for every single client, so instead of rebuilding the same documents from scratch each time, I turned the whole process into a skill. Now I just tell Claude something like "draft an invoice for [client] [amount], [scope]" and it triggers automatically, pulls in my agency details, payment terms and currency, and hands me a clean, ready-to-send document. Same for a letter of agreement. One optimization: watching it run the first time, I noticed it kept asking me for the same details I use on every doc my business info, banking details, payment terms. So I baked those straight into the SKILL.md as fixed context and will keep reiterating, as the lesson mentioned, the more you use the skill, the more you can improve it and the more you can get better. That's very, very important. Now it stops asking and fills them in automatically, and only asks me what's actually different per client. First pass is right almost every time now. The lesson that clicked: a good skill isn't about the AI being clever it's about front-loading everything you already know so you never repeat yourself. A 10-20-minute admin job is now about 30 seconds.

BUILDING AN AI-NATIVE STARTUP Dropping something here that I think every one of you should read: Anthropic's "The Founder's Playbook: Building an AI-Native Startup." Full PDF attached. Why it matters: we're in a moment where someone who has never written a line of code can ship a real product, and the "lean 10-person unicorn" went from a meme to an actual game plan. The playbook maps the entire startup journey — Idea → MVP → Launch → Scale — for a world where AI is your research team, your engineering team and your ops team at the same time. Here's the part that hit me hardest: the biggest danger isn't that you can't build anymore, it's that building is now SO easy you'll skip the part that actually matters. 42% of startups fail because they built something nobody wanted, and AI makes that trap easier to fall into, not harder. This is basically a manual for moving fast without fooling yourself. A taste of what's inside (no spoilers): • Why your job is shifting from doing the work to directing the agents that do it • The three Claude surfaces, Chat, Cowork, Code and exactly when to use each • How to use AI as a devil's advocate to attack your own idea, because your bias now has a research engine behind it • The validation trap: why a working prototype is NOT proof anyone wants your product • "Agentic technical debt": the silent killer that compounds while you're not looking • The real tests for product-market fit (the 40% rule, the push-vs-pull test) • How to build a moat a funded competitor couldn't copy in under two years Real talk though: don't just dump this into Claude and read the 5-line summary. Read the actual thing. Sit with it. The people who take 30 minutes to properly read this will start thinking differently about how they build and they'll quietly pull ahead of everyone who skimmed an AI summary and moved on. That's the whole edge. This applies whether you're building an AI startup, installing an AI operating system for a client, or just trying to work with Claude at a higher level. It's that good.

What I built: A full competitive intelligence + SEO report for one of my clients ahead of their website launch that i built. I used Firecrawl to scrape 6 competitor websites, 198 pages in total then ran the raw data through Claude Code (with Antigravity as my IDE) to turn it into an actual strategy: executive summary, competitor comparison matrix, site audit, SEO action plan, a ready-to-paste content pack, keyword map and a social plan. Then I had Claude Code serve it all as a local interactive dashboard so I could click through it instead of squinting at raw HTML. One thing that clicked: The scrape is just data the real leverage was Claude reading all 198 pages and pulling out the ONE gap none of the competitors fill. That gap is exactly the lane my client can own at launch. Going from a pile of scraped pages to "here's your single biggest opportunity in one sentence" in an afternoon is the part that genuinely got me. One thing I'd improve: Right now this is a one-off, hand-built for this client. Next iteration I want to turn it into a repeatable pipeline, drop in any client + their competitors and generate the same audit on demand, instead of rebuilding it each time. #AISChallenge

What I built: An AI-generated newsletter covering Anthropic's Fable 5 and Mythos 5 launch including the wild twist where the US government ordered both models offline just 72 hours after release. Ran it through the WAT newsletter workflow and kept it short and punchy on purpose. (Screenshots below.) One thing that clicked: How fast you can move now. I went from idea to a working prototype in a single day the kind of thing that used to take me weeks or months. That compression is the real unlock. One thing I'd improve: The voice. Right now it reads close to the default model tone. Next iteration I want to bend it fully into my own style so it sounds like me, not the tool. #AISChallenge