Hello to all. I have been part of this community for a while now and although sometimes (actually most of them) I felt overwhelmed with the amount of information being presented here every single day, I was able to focus on what was helpful for me, for the things I wanted to do. This community really matters. Besides the amazing (and overwhelming) content @Nate Herk shares, there are other folks that are mentors, or at least I adopted them as such, people like @sam-alder-7095 and @Chris Jadama. The reason I am writing about this is because I wanted to share a story of what I would consider a win, which is using the lessons I learn from Nate to add value for people with rare diseases. So here is the story, I hope it't not boring. It started with a tweet. Miriam, a patient with ultra-rare metastatic breast cancer, had accumulated nine years and almost 280 medical documents that nobody had ever seen together. Javi López, a developer, spent a week applying an adversarial AI methodology to her case and shared it openly. I read it and thought: this should exist as a tool. So I built MedSynth. Two AI models debating each other across eight rounds, analyzing a patient's full medical history and converging on structured findings. It worked. But someone with a regulatory background went through the code and showed me exactly where the risk lived. People in desperate situations were going to use this to make medical decisions. It was not ready for that. MedSynth is still paused. Meanwhile, Miriam told me what she actually needed: her 278 PDFs converted into clean, portable, verifiable data. Not a 1000-page PDF nobody can process. Her words: "my data curated so anyone with an LLM project can use them, and not one comma can change." That became the pipeline. Every PDF extracted, anonymized, structured, and verified before being committed to a private repo. Then a wiki layer: an LLM reading all 278 documents and compiling thematic pages covering her full timeline, treatments, lab trends, imaging, clinical trials. Every claim cites its source.

If you're building AI agents, I'd urge you to create a template for prompting. Two notable builds in the last month as proof: - I've built an AI agent that has handled over 9,000 emails - Another AI agent that's handling 25k customers. But here's the full setup: - A claude.md file that references a prompting guideline file, it tells Claude how to write prompts. - Once a prompt is approved, I write at the top "approved for production" which tells Claude that it should not make big changes. This makes sure that the prompt does not get destroyed by Claude. - Push the changes to my GitHub to keep track of all changes. This last part is where most people go wrong. When they see a mistake, they ask Claude to write an explicit rule to never do that again. The issue is that Claude will only look for that exact case, and if the next case doesn't match it, Claude will skip it. Instead, what I do is write mental models of the idea, what we're trying to do and why. When you do it this way, Claude has to use more reasoning to figure out which mental model makes sense. You're letting Claude think with some constraints. But this system has cut down my prompting time and also increased my reliability ten fold. And the thing is that I can use this wherever AI agents are used. Sales agent, customer service agent, any type of agent. Because the structure is the exact same every single time. Give me the agent and I'll make it reliable.

So I've been building a lot with n8n, actually helped 33 clients so far and have a retainer right now. I was thinking of offering a free hour of teaching somebody n8n. Maybe not super basic, but just how to build and make sure things scale. I wanted to check if anybody wants this. Please be serious, don't apply and just ghost, as I'll be teaching and showing everything so you actually learn and can use it.

If you're working with clients right now, tell me if you recognize this. You come across a bug and decide to patch it. That feels logical. But it's a mistake, because you didn't solve the real problem. Before I explain, let me share what happened yesterday. I came across 137+ orders with delivery issues. So I did the most logical thing: sent it over to the CS team and asked what to do. We agreed on a solution for future cases. But here's where I felt dumb. The CMO asked me why it was happening. I had no clue. But that one question flipped my thinking, because now I had to focus on the root cause rather than the symptom. And I bet you're doing the same thing when bugs come up. You see an error, you patch the error. But you never asked why it happened in the first place. That's the real question. Because once you ask it, you stop patching symptoms and start fixing actual problems. And most of the time, the root cause is far easier to fix. In my case, I was about to build 3-4 automations to work around the symptom. One address validator fixed it completely. The lesson: when you hit a bug or an error, ask yourself why it happened and how you can stop it from happening again. You'll notice a stark difference in how you think and operate. And one more thing I realized after asking myself that question: being around the right people is a serious shortcut. That one question from the CMO pushed me to a higher level.

This post: "If you've ever felt 'AI Overwhelm', please read this." Specially this part: Someone can work 12 hours one day and feel insanely productive, but they were just watching tutorials and playing around with new tools. And this: Meanwhile someone else sits down for 5 hours, ships the one thing that actually matters, and makes more progress. I've been in the AI space for 8 months now and once you start getting paid on a project basis, you kinda realize that you have a lot of free time. It's a bit odd at first, but you're in control of your time. And the thing is, if you only work on the main thing that matters, most of the time it's 5 hours of work. This Monday is a perfect example. I worked on a new project for a client and completed it in 3 hours. At that point I could go ahead and chill for the entire day if I wanted to. I should mention that I do lead gen every single day. But out of all the tasks I do day in and day out, these are the most important things: - Lead gen - Complete projects And these things don't take 12 hours. They take 5 to 8 hours, tops. I like working so I do more, but that's because I decide to. So it's quite true: 12 hours on the wrong thing doesn't matter. But on the right thing? Massive difference. PS: this is for another post, but back in February I had 4 projects in the same week and worked the entire week. I earned less on all of those projects combined than what I'm making on 2 clients right now this month.