Notes from
the lab
What the Nodeblue team is working through as we build systems that understand and act inside real operations. The questions we keep pulling on, how we build, and what we learn when the research meets a real floor.
Writing from the lab on how machines can understand and act inside real operations.
All Posts
What We Learned Pointing Nexus at a Real Site
Notes from running our operational intelligence system against real production controllers instead of a simulator. The surprises were not about the model. They were about what diagnosis actually is, and where a system has to hand back to a person.
Dylan McCarthy
June 19, 2026
Building Agents That Act: What the Demo Does Not Show
The gap between a working prototype and a reliable agent running in a real operation is wider than most teams expect. Here is what breaks, and how we design around it when a wrong action has a cost.
Brandon Sheedy
June 10, 2026
Evaluating Systems That Act in the Real World
A benchmark score tells you how a model does on a fixed test. It tells you almost nothing about how a system will behave inside a live operation. Here is how we think about evaluation when the output is an action, not an answer.
Robert Anspach III
June 1, 2026
Why Dashboards Stopped Being Enough
The dashboard was a genuine advance. It is also where a lot of operations got stuck. Observation is not understanding, and adding another chart does not close the gap between them.
Dylan McCarthy
May 22, 2026
How Operational Knowledge Decays
Most of what an organization understands about itself lives in people, and it leaves when they do. Documentation goes stale the day it is written. We study what it would take for a system to hold that knowledge and keep it current.
Brandon Sheedy
May 13, 2026
Judgment and Recall: Dividing the Work Between People and AI
The useful question is rarely whether a system can do a task. It is which parts of skilled work should be handed to a system, which must stay with a person, and how the seam between them earns trust.
Robert Anspach III
May 4, 2026
The Gap Between Advising and Acting
Describing a situation is one thing. Acting on it safely is another. The distance between the two is where most applied AI quietly stops, because in the physical world a wrong action has a real cost.
Brandon Sheedy
April 25, 2026
Forge: Turning Engineering Intent Into Working Systems
A look at the research behind Forge, our effort to let teams describe what a system should do and get a trustworthy first version of the logic and configuration built with them, rather than from scratch.
Brandon Sheedy
April 16, 2026
AI for PLC Troubleshooting: What Works on a Live Floor
Most AI troubleshooting tools were designed for a demo, not a down line. What fault diagnosis actually requires, where AI genuinely helps, and the deployment constraints a real plant imposes.
Dylan McCarthy
April 4, 2026
Generating PLC Documentation with AI, Graded by Controls Engineers
Machine-written PLC documentation is only useful if every statement traces to the rung that makes it true. How we generate program docs, and how 137 of them survived grading by three independent Controls Engineers.
Dylan McCarthy
March 24, 2026
Can AI Read PLC Code? An Honest Assessment
Large language models can discuss ladder logic convincingly and still get the rung wrong. What it actually takes for AI to read Studio 5000, TIA Portal, CODESYS, and the rest of a plant's control logic correctly.
Dylan McCarthy
March 12, 2026
Following the research?
We publish what we learn as we build - from the research direction to field notes from working deployments.