Context Engineering in Industrial AI | SPREAD

The models are not the bottleneck anymore. In his opening at CUBE 2026, SPREAD co-founder Robert Goebel argued that the edge in industrial AI has moved to context engineering: the connected, typed product knowledge an AI system needs to reason correctly about a physical product, with time and provenance built in.

A year ago, betting on context over raw model size was a contested position. For SPREAD it was the founding bet back in 2019. At CUBE 2026 it was close to industry consensus. The talk lays out why the constraint in engineering is not the model's reasoning, but the absence of a connected model of the product for it to reason over.

In his opening, Robert covers:

• Why context, not compute, is the real constraint in industrial AI
• The three prerequisites of engineering context: a typed map of the system, with time and provenance built in
• Why connecting data pilot by pilot builds "context islands" that can never be joined back together
• What it means to treat context as shared infrastructure instead of a scatter of experiments

Watch the full opening above. To go deeper, see how SPREAD builds a connected model of your product on the SPREAD platform and read our guide to what Engineering Intelligence is.

CUBE 2026 brought senior engineering leaders together across automotive, defense, and cloud. See more from the day: Philipp Noll on the agent-native product lifecycle, the SPREAD and Salesforce integration for connected vehicle diagnostics, and the full CUBE 2026 recap.