Arcgate

Another step forward in my software engineering journey completed the Docker Foundations Professional Certificate. Learned core concepts of containerization, deployment workflows, networking, and working with Docker across multiple cloud platforms and modern development environments. Looking forward to applying these skills in real-world projects. #Docker #DevOps #CloudComputing #SoftwareEngineering

The first AI feature I built didn't work. Here's why. And what I learned. Last month I decided to build something with LLMs. Seemed straightforward. A feature that would take user input, run it through Claude API, return an enhanced version. Simple prompt. Solid use case. Should work. I built it in a weekend. Then I tested it with real data. And immediately saw the problem. The API was returning different outputs every single time. Same input. Different result. Every call. I thought I was doing something wrong. So I debugged the code. Three times. It was fine. Then I realized: the problem wasn't my code. It was my thinking about how LLMs actually work. I was treating the LLM like a traditional API. Input → Output → Done. Reliable. Deterministic. Predictable. LLMs aren't like that. They're probabilistic. They're creative. They're inherently inconsistent. That's their strength in some cases. It's a disaster in others. I was building a feature where consistency mattered. And I had chosen a tool that's fundamentally inconsistent. The fix wasn't in the code. It was rethinking the entire approach. Lower temperature settings. More specific system prompts. Different model choice. Validation layers. Fallback logic. All things that would have been obvious if I had understood what I was actually building with. Here's what I learned: Most tutorials teach you the syntax. How to call the API. How to structure a prompt. What they skip is the deeper question: What are you actually building with? Not what it can do in a demo. What it is: philosophically, technically, practically. LLMs are powerful at generating creative content, finding patterns, reasoning through ambiguity. They're terrible at consistency, factual accuracy, and deterministic outputs. If you're building on top of that foundation without understanding it, you're going to hit walls that code can't fix. I'm paying attention now in a way I wasn't before. And it's changed how I think about what to build and how. What's an LLM feature you tried to build that didn't work out? I'm curious what other people ran into. #AI #SoftwareEngineering #GenerativeAI #TechTrends #ProductDevelopment #BuildInPublic

There's a question every developer eventually asks themselves. Not out loud. Quietly. Usually late at night. "Am I building a career, or just a resume?" For a long time I didn't have an answer. I had skills. Real ones. I had experience. Legitimate experience. 4 years of shipping systems that worked, at scale, under pressure. But somewhere along the way I started noticing something. The most satisfied developers I knew weren't just good at their craft. They had something beyond the job. Something they were working toward. Something that was theirs. I've been thinking about that a lot lately. What it actually looks like to work a full-time job and simultaneously build something of your own. The discipline it requires. The trade-offs it costs you. The specific kind of loneliness that comes from working hard on something nobody around you knows exists yet. And eventually, the moment when it all becomes visible. I don't know when that moment is. But I'm building toward it. Every single day. This page is where I'll document that journey. Not a highlight reel. Not a success story, at least not yet. Just the honest, unfiltered version of what it looks like to be a developer who decided that building things for other people isn't the whole story. Are you building something beyond your job right now?