Hello everyone, I've completed my first task of building a newsletter using Claude code. Since it was my first time working with Claude code, I wanted to keep things simple by following the first lesson by Nate. Everything executed perfectly without any errors for this task. For the second day, I plan to try something on my own while still following the lessons for the upcoming days. I would really appreciate it if anyone could share their use cases. Thank you!

Like most people, I'd been using ChatGPT to write things, ask questions and review the odd bit of code, and I thought that meant I was using AI. It didn't. Not even close. Then I stumbled onto Nate's channel (and now this community), and it honestly changed the way we work. The WAT framework was probably the biggest lightbulb moment. Nate explains agentic workflows in a way that just makes sense. I watched the 36-minute Claude Code beginner video three times. Not because I had to, but because every time I watched it, something else clicked. Since then we've completely changed how we build software. We now have agents with specific jobs, memory that carries context between sessions, reusable skills, and workflows that mean we're not constantly repeating ourselves. Instead of spending hours on the boring repetitive stuff, we let the agents handle it and focus on the parts that actually need us. The time savings have been insane. Projects we'd normally estimate in months are now taking weeks. But the biggest change wasn't even the speed. I actually enjoy working again. There were parts of this job I used to dread because they were repetitive or just mentally exhausting. Now I catch myself looking forward to opening my laptop in the morning. I honestly didn't expect AI to do that. Massive thanks to Nate for everything he shares. And for those just checking in, wondering whether it's worth learning... it is.

Completed day 2 and glad I finally got to see what Firecrawl can do. I always heard it can scrape sites but hadn't seen any use cases. Had a hiccup along the way with connecting to firecrawl's MCP server from Claude Code… think in the video Nate used the 'claude code' section command but that didn't work for me. It worked when I used the 'VS Studio' link on the Firecrawl website to connect to MCP (maybe it's changed since the video was made). 2nd small hiccup was when my Firecrawl account ran out of credits while Claude Code was running the prompt to scrape a job site to a csv file. It just stopped with less listings so I still got to see the final product which was cool (this task took 1000+ credits 🤷‍♀️). I used my work email to sign up for more credits to run the next use cases which only took 1 credit each haha. Both created nice looking artifacts below, analyzing Myclaw's brand colors and Blue Bottle's site mapping which was impressive!

Diving Deep into the WAT Framework, Token Optimization, and CLAUDE.md Just wrapped up Day 1! Instead of just mindlessly running the setup, I had a great discussion with Claude about the core architecture and the role of CLAUDE.md. It gave me some really solid "aha!" moments about how AI agents actually scale, so I wanted to share my takeaways with the community. 📝 What I Built - The Deliverable: A fully automated newsletter system that takes a single prompt, conducts research via Perplexity, generates visuals with Nano Banana, formats everything in HTML, and sends a polished email via Gmail. - The Setup: Configured the project structure based on the WAT framework (Workflows + Agent + Tools) driven by CLAUDE.md. 💡 My "Aha!" Moments & Core Takeaways 1. What does CLAUDE.md actually do? I realized that CLAUDE.md is the literal "brain transplant" (system prompt) that turns a generic, jack-of-all-trades AI into a highly structured, autonomous operations manager for this specific project. It strictly enforces: - Role Limitation: Stopping Claude from doing everything probabilistically and forcing it to delegate to deterministic Python scripts (tools). - Shared Context: Drilling in the exact folder architecture (workflows/, tools/) and setting ground rules, like keeping deliverables in cloud services. - The Self-Improvement Loop: Commanding the agent to not just report errors, but to actively fix the scripts and update the markdown documentation on the fly. 2. The Dilemma of Centralized Management vs. Token Bloat While I loved the idea of keeping all workflows and tools in one project folder for reusability, a critical question hit me: "As we add dozens of workflows, won't Claude read everything every time and absolutely explode our token consumption?" 【How it's solved】 I learned that robust systems don't force the AI to read every single markdown file from the jump. Instead, they use a lightweight index file or vector search (RAG) to let the system pull only the specific workflow required for the task. It was eye-opening to see how decoupling AI reasoning from deterministic system control keeps the context window clean and scalable.

Hey everyone! 👋 Sharing my Day 1 build — got my newsletter pipeline up and running with Claude Code, and honestly the result surprised me. The moment things actually started working end-to-end was a good feeling. Still have rough edges and things I'd do differently, but as a first iteration I'm proud of where it landed !! One thing I'd improve: the workflows need more refinement.. Every run teaches you something, and I want the next version to be sharper from the start. Would love any feedback from those of you who've been at this longer — always looking to level up. 🙌 #AISChallenge