Become a .NET & Blazor expert with weekly tutorials featuring best practices and the latest improvements, right in your inbox.
I told an AI I’d see it in 7 hours. It finished in 2.
|
Hey friend, Lately I told an AI I’d see it in 7 hours, and went to bed. It was done in 2. Across two different projects. There’s this thing people call the night shift. You give an AI agent a pile of work, start it in a loop, and go to sleep while it keeps working on your real codebase. I tried it a month ago and could not make it work, the model just fell apart. But with the newest models, it works now. And the surprising part wasn’t that it ran all night. It’s that it ran out of work before the night was over. My latest video breaks down exactly what happened, the actual prompt I gave it, the one command that makes the loop run on its own, and the honest look at where this is genuinely useful and where it isn’t. The short version: I asked it to build a performance profiling feature and then actually use it to find and fix slow pages, plus add loading skeletons everywhere. I ran it on two projects at once, mine and a ticketing system I’m building. I woke up to two working apps, all committed in steps I could review. Then I checked the timestamps. The last one was around 2am. It hadn’t run out of steam. It finished everything I gave it and stopped. Which flips the whole question. It’s not “can the AI last all night,” it clearly can. It’s “how much well-defined work can I actually line up for it.” In the video I get into the real reason this works now, why the weekend might be the better move than overnight, and the one thing that made it safe to let an agent run loose across two codebases while I slept, which has nothing to do with the AI and everything to do with how the projects are built. Take care, Patrick P.S. The reason I could trust an agent overnight comes down to clean, AI-ready project structure. That’s exactly what I cover in my free Blazor AI Jumpstart course, the setup and workflow I wish I’d had when I started. |
.NET Web Academy
Become a .NET & Blazor expert with weekly tutorials featuring best practices and the latest improvements, right in your inbox.
Hey friend, A developer I know built his whole career on one bet: that young developers would never want to touch old code. So the legacy systems quietly running real companies would always need someone, and he would be that someone. For twenty years it paid. Then AI arrived and went straight for the part he thought was safe. Because the hardest part of modernizing an old system was never the typing. It was the understanding. Reconstructing thirty projects and three contractors' worth of...
Hey friend, The other day I found a bug in StackBuddy, an app I built. Normally I'd open it, file a ticket for myself, and move on. Instead I turned to my AI and just said, "log a dev ticket for that." And it did. A real ticket in my own bug tracker. Title, description, priority, the whole thing. I never opened the actual app. The thing that makes that possible is called MCP, Model Context Protocol. Ignore the name for a second. Out of the box, an AI can only talk. It writes and explains, but...
Hey friend, I don't really write code by hand anymore. I describe what I want, and the AI builds it. For a while, that scared me. If I am not writing the code, does everything I spent years learning still matter? Does my stack matter at all? If you are a developer right now, I would bet you have felt some version of that too. Here is where I landed. It flipped the whole thing for me. The other day I was debugging a slow app, talking it through with the AI. Would WebAssembly help here? Is it...