Deep Dive AI • Real-World Agentic AI I Let an AI Agent Find the Best Pizza Deal — Then It Actually Ordered Dinner There are moments with AI when something ordinary suddenly makes the future feel very close. Mine happened while trying to buy pizza. REAL-WORLD TEST The goal was simple: find the best pizza deal. I was not sitting down to build an automation demo. I was not testing a benchmark, writing code, or trying to manufacture an impressive AI story. I wanted dinner. Normally that means opening search results, checking restaurant specials, bouncing between menus, looking for coupons, comparing totals, choosing a pickup time, typing everything into another website, and then keeping track of the confirmation. This time I handed the goal to an AI agent instead: scour the pizza deals around Charlotte, Michigan, identify the best value, and help get the order placed. The important part is what happened next. It did not merely return a list of pizza restaurants and ...

Michigan MRI · Field guide How to Use MRI CON Assist Find the right referring provider, inspect the source license, and copy the Michigan MRI CON value with a deliberate verification step. Open MRI CON Assist ↗ Open MRI Safety Navigator ↗ Two independent MRI workflow tools. For provider codes, start with MRI CON Assist. You are working at an MRI console when you need the referring or ordering provider’s license value for Michigan MRI Certificate of Need reporting. The name may be common, the license may use a different format, and the profession matters. One wrong match can turn a ten-second lookup into a scavenger hunt. That narrow problem is why we built MRI CON Assist . The app supports one careful workflow: search, review every plausible match, select the right provider, inspect the source license, generate the configured 10-character value, and copy it. Nothing more. No mini-RIS hiding behind a cheerful button. What the App Does Toda...

Friday in Frankfort: E-Bikes, a Low Battery, and the Perfect Stormcloud Stop We came into Friday with a bigger ride in mind. The plan was roughly 20 miles on the e-bikes, starting from Crystal Lake RV Campground outside Frankfort and using the Betsie Valley Trail to explore the area. Northern Michigan gave us exactly the kind of riding we came for: cool September air, woods beginning to hint at fall, water nearby, and long stretches where the bikes could simply roll. Then the battery gauge had other ideas. At about 7 miles out, my battery was already showing a low charge. We could have kept pushing toward Beulah, but that would have turned a relaxed vacation ride into a very different kind of adventure. So we turned around. That left us with roughly 14 miles for the day, which turned out to be plenty. Sometimes the best travel plans are the ones you're willing to change before they become bad stories. And besides, Frankfort was waiting. Post-Ride Recovery, Northern Michigan...

Day Two at Camp: Pie-Iron Breakfast, Cascading Tomatoes, and an E-Bike Ride to Beulah Friday morning started with exactly the kind of breakfast that makes camp cooking worth the effort. Yesterday, our fresh tomatoes helped build the base of the Dutch oven chili. This morning, those same tomatoes changed jobs. That is the whole point of this cascading camp menu: ingredients keep moving, but the meals do not feel like repeats. Today, the tomatoes went into bacon, egg, cheddar, and fresh tomato pie-iron melts. And judging by the finished sandwich, the pie iron did its job. The bread came out deeply toasted and compressed, with those unmistakable ridges across the top. Cheese and bacon escaped around the edges just enough to remind us that camp cooking is not really about perfect presentation. It is about whether the cheese melted. It did. Friday Morning: Bacon, Egg, Cheddar & Fresh Tomato Pie-Iron Melts The basic recipe is simple: thick slices of bread eggs bacon s...

First Night at Camp: Dutch Oven Chili, Goldfish Crackers, and Late-Night Cornbread Thursday night was exactly what the first night of a camping trip should feel like. We were finally settled in. The fire was going. The cooler was unpacked. The camp kitchen had officially taken over. And dinner was already waiting for us in the plan: a big Dutch oven of late-summer chili built around fresh tomatoes, ground beef, beans, peppers, garlic, and enough seasoning to make the whole campsite smell like dinner before the bowls were even filled. This was also the first real test of our cascading camping menu. The idea for this trip was simple: cook ingredients once, then let them change jobs over the next few days. We did not want four completely separate menus, four separate grocery lists, or a weekend built around eating the exact same leftovers repeatedly. Thursday's chili was designed to start that process. It would be dinner tonight, and part of it would eventually become Friday...

My AI Factory Actually Works. Now Comes the Hard Part. For months, I have been building what I somewhat grandly call the AI Factory. The idea sounds simple when I reduce it to one sentence: Give an AI system a piece of content and let it help turn that into a finished, validated, publishable media project. Of course, that sentence leaves out approximately 97% of the engineering. There are transcripts. Media files. Metadata. AI-generated visuals. State machines. approval gates. OAuth. YouTube APIs. Blogger. Local models. Python. PowerShell. GitHub. Multiple repositories. Recovery logic. File hashes. Production checkpoints. And, because apparently I enjoy creating problems for future Jason, more than one AI agent involved in the process. Over the last several weeks I have been working heavily on the Factory's Hermes integration: a locally running AI supervisor that can inspect projects, understand state, invoke carefully limited capabilities, and eventually supervise ...

I Spent an Hour Just Starting My AI Factory — So I Built Hermes a Start Button The problem was not that the AI Factory had no tools. The problem was that every restart felt like rebuilding the control room from scratch. This morning I sat down at my home PC with what should have been a very simple goal: Turn the AI Factory back on and continue the video that was already sitting halfway through production. Not start a new project. Not rewrite the pipeline. Not redesign Hermes. Not debug OpenMontage. Just continue yesterday's work. Instead, nearly an hour disappeared into ports, model blobs, YAML configuration, local inference servers, Hermes profiles, MCP tools, project folders, checkpoints, logs, and one question that became increasingly difficult to ignore: Why does a system this sophisticated still need a human to remember how to start it? That question led to one of the least glamorous — and potentially most important — additions to my local AI Factory so fa...