Field notes from inside the work. Long-form writing on cognitive load, drift, and the future of MEP project management.
The industry has seen versions of this before. This time, our contracts, our schedules, our production history, and thirty-year careers are the training data. That changes the question. For years we asked who owns our data. The bigger question is becoming who owns the intelligence created from it.
Read the full pieceAI allows the value created through labor to outlive the labor itself. Artificial intelligence can process information. It cannot manufacture experience. Construction is not simply another customer. Construction is the teacher. For the first time in decades, construction is negotiating from a position where it possesses the scarce resource. That is leverage.
ReadThe next 10 years can be the story of an industry taking ownership of its institutional knowledge. Or it can be the story of watching that knowledge and the value of its labor get stripped away. Some capture your knowledge, structure it, and rent it back to you. Your teams generate the intelligence. Their shareholders keep the upside. The trades cannot afford to be next.
ReadSub-$100M shops are about to have their insurance company tell them what tech to use. The logic behind sequencing, load calculations, and trade coordination has been refined over thousands of years. Construction should be driving this. Not riding in the back seat. And definitely not on the menu.
ReadI've gone quiet for a while. Heads down building and testing DPM, and bringing on our advisory board. It's been bananas. So we did what you do with bananas. Turned them into banana bread, Cheesecake Factory sized, with the ROI to match. Here's the deal: everyone except our advisory board is getting cut.
ReadRecent government restrictions on advanced AI models made one thing clear: the frontier model you depend on can get locked down with little warning. We saw this coming years ago. That shaped how we think about building Documint PM: around flexibility and data ownership. Your shop's knowledge should stay yours.
ReadOur industry uses data, information, and intelligence almost interchangeably. Dashboards promise intelligence. AI promises intelligence. Reporting tools promise intelligence. Most of what they deliver is organized information. Those are not the same thing, and the gap between them is where cost and schedule impacts actually live.
ReadAfter months of building, testing, breaking things, fixing them, and testing again, my perspective has changed. I still believe now is the time to start. Just not for the reason I originally thought. Don't start because you're afraid of missing out. Start because now is the right time to learn. Especially in MEP construction.
ReadData that arrives after the decision is not data. It is a receipt. In 2021, Autodesk and FMI estimated that bad data may have cost the global construction industry $1.85 trillion in a single year. The next generation of construction technology will not win because it stores more information.
ReadEvery meaningful advancement in the MEP trades has been met with resistance. PVC solvent cement was going to ruin plumbing. ProPress couldn't be trusted. BIM was a distraction. There is another burden building on projects right now, and we are not measuring it.
ReadThree months running Gemma 4:31B locally on a Jetson AGX Orin 64GB, tied into Supabase, image analysis, web, and Telegram. The side project turned into the future. The companies that build internal AI memory first are going to have a massive advantage.
ReadIt's 1:47 PM Pacific. There's a piece of equipment that needs to be ordered today, not tomorrow. Every PM in the MEP trades has lived some version of this day. What nobody in this industry has done is count it. SDE → DRV → ADS → CC, a framework for the load no one is measuring.
ReadOne long-form piece a month. No marketing. Field-tested writing on what's actually moving in the MEP trades.
Subscribe