I keep going back and forth between: “I should learn a bit more first” and“ I should just start and figure it out”. Some days I feel like I’m close to understanding it, other days it feels like I’m missing something obvious. Not sure if that’s just part of the process or I’m overthinking it. Anyone else went through this at the start?

Hey everyone 👋 I’ve been learning a lot here around AI workflows, systems, memory, and modular design. Instead of just consuming, I want to start building more seriously. So here’s a challenge: Give me a real problem or use case, and I’ll try to build an AI system for it. It can be anything: - automation - agent-based workflow - memory system - business use case I’ll build it, test it, and share what works (and what doesn’t). Trying to move from learning → building. Let’s see what we can create 🚀

@Alex Nartey this one is for you buddy. After reading your thoughts and chatting with you I realized showing rather than telling you about taking action just makes more sense. ---------- The SMS Bridge project was born out of a client (JD) a marketing agency asking 1 question could we improve or automate their SMS marketing operation using their existing. V1 was ugly and clearly vibe coded from the look and feel to the sql light data base claude code picked without context, yeah I had to have it rebuild the DB. After fixing the db issue in v1.5, then I had a something that I could test with JD in less than 20 hours of effort. So we got going in addressing JD's unique pain point and that's where I was able to start scoping out a Product Requirements Document for what became V2. V2 was built to scale up or down with how JD saw their business growing. JD ran this as inhouse software with inhouse hardware (as that's how they wanted things to run). It was my first build using AI to help and I made a ton of mistakes, didn't understand the power of having a folder structure, a good IDE, containerizing things, branching best practices, etc. That stuff and more all matters. What matters more is ACTION not reckless (do some research, I did) BUT don't get stuck on reading for ever make the thing even if it's a 'localhost' desktop app, then learn your next step.

Hey everyone 👋 As I’ve been going through discussions here, I’m starting to notice a pattern: There’s a big difference between: - simple workflows - and actual AI systems At first, I was mostly building workflows (input → LLM → output) But now it feels like real systems involve: - structure (context, modular parts) - memory (logs, learnings) - feedback loops (improving over time) - multiple roles instead of a single call So instead of just calling an LLM, you're designing how the system thinks and evolves. Still learning, but this shift in thinking feels important. Curious — what do you think is the biggest difference between a workflow and a real AI system?

Hey everyone 👋 I joined this community recently and have been actively going through posts and discussions. One thing I’ve already learned is that structure matters more than just prompts. Before this, I was mostly focused on getting better outputs, but now I’m starting to understand the importance of context, memory, and workflow design. Even small changes in structure can make a big difference. Still learning, but this shift in thinking is already helping me a lot. Curious — what was your biggest “aha” moment when learning AI workflows?