Dyad 3.2: More LLMs and Better Visualization
The latest Dyad 3.2 update significantly expands the platform's capabilities by broadening the AI agent's model selection and improving the graphical workspace. Users can now select OpenAI models like GPT-5 (featuring a 1M-token context window and adjustable reasoning effort) alongside Anthropic models, backed by improved configuration reliability and API error diagnostics. The graphical editor makes building large models easier by visualizing analysis points directly on the canvas and fanning out array connectors into individually wireable sub-connectors. Rounded out by direct parser-based syntax highlighting and a seamless, one-click migration path for existing 3.1 projects, Dyad 3.2 delivers a more powerful and intuitive modeling experience. Find out more.
Discovering Missing Physics: Why Your CAE Models Fall Short (and How the Dyad Agent Fixes it)
Traditional Computer-Aided Engineering (CAE) models often struggle to capture the complex, non-linear behavior of real-world systems, while pure AI models can sacrifice physical accuracy and interpretability. In this technical brief, discover how Scientific Machine Learning (SciML) combines first-principles physics with neural networks to uncover "missing physics" and build high-fidelity digital twins. Through real-world automotive case studies, see how the Dyad Agent simplifies the entire workflow—from model selection and training to optimization—using natural language, helping engineers develop more accurate simulation models faster. Read the blog and download the technical brief to learn more.
Upcoming Webinars & Events
Agent Simulation of the Activated Sludge Process (July 22)
Running a wastewater treatment plant requires a delicate, manual balancing act between high energy costs and strict effluent quality standards—a challenge the Dyad Agent transforms by using natural language to simulate, interrogate, and optimize an activated sludge model in real time. In our upcoming webinar, you will watch the Agent automatically execute parameter sweeps across aeration rates to map the exact sweet spot between energy consumption and regulatory compliance. This powerful workflow—modify, simulate, sweep, and reevaluate—eliminates the need for deep modeling expertise and provides a universal blueprint for engineers looking to optimize virtually any complex physical system.
Autonomous HL‑20 Model Creation (July 29, 2026)
Join JuliaHub on July 29 at 12:00 PM EDT for a live webinar demonstrating how the Dyad Agent autonomously transforms dense technical specifications into validated, physics-enforced simulation models without manual coding. Using NASA’s heavily documented HL-20 Personnel Launch System as a complex, multi-domain stress test, speakers David Dinh and Rajeev Shobhit Voleti will show how the agent reasons over aerodynamics, inertias, and control laws, self-corrects via a compiler-backed constraint system, and validates the final results against NASA's published trim and dynamic data. Attendees will learn how this transformative AI capability extends beyond aerospace to help any engineer rapidly move from specification documents to verified simulations.
Insights from the Lab
Why Early Adopters Move to Dyad
Author: Dr. Michael Hoffmann
After 50 days of deep-diving with customers, Dr. Hoffmann explores the "why" behind the jump to Dyad. It’s more than just a faster compiler; it’s a fundamental shift in how complex systems are built and validated.
One Model, Two Analyses: Sizing vs. Transient Simulation
Author: Dr. Ranjan Anantharaman
Stop maintaining separate models for steady-state sizing and transient dynamics. Dr. Anantharaman demonstrates how a single source of truth in Dyad ensures consistency from equipment specification to 24-hour operational analysis.
Rethinking Physical Modeling: Moving Beyond Blocks and Math
In this blog, Dr. Michael Tiller explains how the traditional way we model physical systems is fundamentally inadequate, how acausal modeling changes the game, and why the Dyad platform represents the future of engineering design.
