Modelware founder and CEO, Dr. Maged Elaasar, will present the paper Beyond Syntax: Method-Compliant AI Assistance for Low-Resource Modeling Languages, co-authored with Sokhna Amar and Sambit Bhattacharya, at MDE Intelligence 2026, the 8th Workshop on Artificial Intelligence and Model-driven Engineering. The workshop is co-located with MODELS 2026 in Málaga, Spain, in October 2026.

Coding assistants are trained on the languages the internet is full of. Modeling languages are not those languages: the public corpus is thin, and an assistant that has only seen a little of one tends to produce output that parses but does not follow the method behind the language. Syntactic correctness is the easy half of the problem.

The paper takes up the harder half, and the question of how to measure it. It presents an evaluation framework and a protocol-driven architecture in which the language model interprets intent while every edit is routed through validated MCP tools, the same write operations the OML Code IDE uses, with method compliance enforced by a SHACL shape catalog encoding the rules of the Sierra systems engineering methodology, a DSL defined within OML. Tasks are drawn from the public FireForce VI aerospace case study across three difficulty tiers and scored on syntax accuracy, method compliance, and API cost.

The headline finding: tool augmentation helps most at the higher difficulty levels, which is to say deterministic scaffolding matters most exactly where probabilistic generation is least reliable. We wrote about what that means in practice in Syntax Is the Easy Half.

Accepted workshop papers are published by ACM in the joint MODELS workshop proceedings.

Attending MODELS 2026? Reach out to meet with us in Málaga.