Hey Delivery Pilot family 👋 Quick reminder: our Sunday Live Cohort is tomorrow at 18:00 UK. If you haven’t made it to a recent Sunday yet, this is a great one to jump in. You don’t need to arrive with a polished plan — come for live help, screen sharing, and building with others instead of figuring it out alone. What we usually do in these sessions: • Interactive screen sharing and live troubleshooting • Walking through the week’s certification-prep video together • Building real projects (Second Brains, predictive models, serverless workflows) with peer feedback • Optional screen shares with respectful turn-taking — share when you’re ready Catch up on previous cohorts / recordings so you feel ready: • Cohorts hub: https://www.skool.com/delivery-pilot-8938/cohorts • Cohort 11 – Value AI & Course Creation Incentive: https://www.skool.com/delivery-pilot-8938/classroom/781f28c5?md=24745fd3743f40df835065169393ecd2 • Cohort 7 – Second Brain: https://www.skool.com/delivery-pilot-8938/classroom/781f28c5?md=cd4f630c20af4291b5f19ae78fc5ca1b • Cohort 6 – Forward Deployed Engineering: https://www.skool.com/delivery-pilot-8938/classroom/781f28c5?md=fcbf7ebf37bd4e62817c9f0cbb1472af • Cohort 1 – Foundations: AI Market & Certification: https://www.skool.com/delivery-pilot-8938/classroom/781f28c5?md=60bb8f6b3d244b74abda6f4beb95d628 See you tomorrow at 18:00 UK — excited to build with you 🙌

Courses Production Hey everyone! 👋✨ Full transparency: we’re building this engine while flying it, and we are loving every second of doing it with you! 🚀 Big love and appreciation for all your feedback, ideas, and collaborative energy—this community is truly shaping what this space becomes, and we couldn’t be more hyped to build it together. 🤝💡 While we’re hard at work crafting the ultimate experience behind the scenes, we want to make sure you get exactly what you need right now. Here is where you can dive in depending on what you're looking for: 🎓 Want the Official Basics? If you're looking for the official setup straight from the source, check out: 👉 Anthropic’s Official Claude 101 Course 🔥 Prepping for Exams & Building Together? If you want focused exam prep and want to learn, practice, and build alongside the rest of the crew, drop into our unofficial community hub: 👉 Skool Exam Prep Classroom 🎬 Watch & Review the Videos in Canva All course videos are hosted inside Canva, and we need your input directly in the Canva comments section! We’re using a multimedia learning framework designed to spark self-driven learning. When you analyze the videos from both an operational (process, mechanics, execution) and artistic (flow, visual impact, engagement) perspective, you help shape the content. Dive into the comments and let's hit these learning objectives together! 💡🎨 Interested in the production pipeline Check > https://rifaterdemsahin.github.io/VideoProductionPipeline/index.html Keep the feedback coming, jump into the discussions, and let’s keep building! 🙌 🎉

AI Related job postings > Feel free to post the jobs post that are interesting in AI. https://ai-job-postings.polished-boat-17b2.workers.dev

There are some interesting webinars from Antropich🥳I highly recommend you to join them, with so, we can discuss and insipire each other during our sesssions. I've registered the first upcoming one, (Maximizing value from claude code) which is on 10th September morning. Adding the link below for your interest🤗 https://academy.claude.com/webinars

In RAG, how you split a document into chunks directly determines the quality of the system, but there's no clear answer to "which strategy is best," it's usually chosen by intuition. So I built a tool that makes this measurable. Grain splits the same document with 6 different chunking strategies, tests each one with real embeddings, and shows which one is both the most accurate and the most efficient (fewest tokens). 🔗 Live: https://grain-cdw.pages.dev What I learned along the way, the part I most want to share with you: - My first measurement method turned out to unfairly favor small chunks without me realizing it. I tried 4 different approaches to fix it, moving to the next one whenever one didn't work. In the end, the root cause wasn't in the metric itself, it was in the retrieval protocol. - I kept challenging myself with "does this actually work?" Something that passed with synthetic test data failed a few times on the first real API run. Every failure was a real engineering lesson. - Managing a project at this scale with Claude Code end-to-end (from architectural decisions to production deploy) gave me a real chance to apply the discipline I've been learning here: typed error handling, avoiding the "right concept, wrong layer" trap, never deciding without evidence. If you want to try the product, it's at the link, you can upload your own document and see the real result. Always open to feedback. 🙏