Frequently asked
questions.

Nodeblue is an independent applied research lab studying how machines can understand and act inside the operations the real world runs on. We are defined by a research direction, not a product line. We build systems to find out where the ideas hold and where they break.

Nexus is the system our research runs on today. It is operational intelligence for industrial environments: it reads a site's control logic, supervisory data, and documentation together, reasons over them as one connected picture, and answers operational questions in plain language, cited to the exact rung. It is the first place our work runs on a real floor.

Yes. Nexus is released in stages, from the deterministic engine to a connected view of the whole operation, and each stage is available now. The hosted product demo is locked while we onboard by hand, so the way in is to request access or book a walkthrough and we will run Nexus against a project like yours.

Studio 5000 and RSLogix 5000, Siemens TIA Portal and STEP 7, Ignition, and the CODESYS family directly, plus one PLCopen parser that reaches 500+ OEM controller brands. It is cross-vendor by design, so adding a platform is a module, not a rewrite. It also reads device configs, manuals, specs, history, and systems of record like MES, ERP, and CMMS.

You ask Nexus a question in plain English and it answers, so the surface looks familiar. The difference is underneath. Nexus is a deterministic engine that parses the logic exactly, reads the live controller, and cites the rung it used, so the same question comes back with the same answer every run. A model pointed at a raw export has to guess at structure it cannot see, and it guesses confidently. We tested eleven frontier models on which controller edit was actually running live and every one got it wrong. With Nexus underneath, all eleven got it right. The model on top is replaceable, and Nexus gets better as those models improve.

Nexus runs on premise and is air-gap capable by default. It holds a persistent, on-premise memory of your operation and never sends your logic or data out to do its work. This is the deployment shape defense, utilities, and pharma actually require.

Yes, on real production data rather than synthetic demos. It has parsed 4,386 real PLC project files from live operations, spanning Rockwell, Siemens, Ignition, and the CODESYS family, with zero parser errors. Three independent Controls Engineers graded 137 of 137 AI-written program documents as rung-accurate with zero errors found.

The early connectors our engine was built on are public on GitHub under the MIT license, so you can see exactly how SCADA access, PLC parsing, and cross-system correlation are handled. The tooling we build today stays in-house, but those open connectors are how we show our work to engineers who do not trust black boxes.

Four directions: operational intelligence (the work behind Nexus), knowledge and memory (keeping what an organization knows), human and AI collaboration (how people and systems share the work), and autonomy and execution (what lets AI move safely from analysis to action). Two early systems, Forge and Atlas, are in active development.

A small group of researchers and engineers with backgrounds at companies such as Tesla, Amazon, and Applied Materials. We hire people who have built real systems in environments where being wrong has a cost, and we prioritize depth over breadth and evidence over opinion.

We are based in St. Petersburg, Florida, and work with industrial operations across the United States. Because diagnosis on a live floor is physical, on-premise work, we deploy on-site wherever the equipment is.

Request access to Nexus or book an intro walkthrough and we will run it against a project like yours. You can also reach us at [email protected]. The fastest way to see whether this fits your operation is to put it on your own controllers.