/

/

November 2026 Newsletter: Dyad 3.5 and Open-Weight Models for Physical AI

November 2026 Newsletter: Dyad 3.5 and Open-Weight Models for Physical AI

November 2026 Newsletter: Dyad 3.5 and Open-Weight Models for Physical AI

Dyad 3.5: Synchronous Control Meets Physical Modeling: Dyad 3.5 adds a synchronous compiler. You can now model discrete controllers and other discrete-time logic alongside continuous-time physics, then export the controller as C code for embedded hardware. What’s new:

  • MPCComponents: a new library for model predictive control

  • DiscreteComponents 0.6: adds a scheduled discrete PID

  • DyadControlSystems 2.6: adds discrete PID autotuning

  • Facade Editor: a new editor for component icons

  • Julia 1.13.2: Dyad now runs on it, with MultibodyComponents bundled

  • Dyad AI Agent: now runs through the Dyad CLI, keeps its Julia session across compiles and defaults to Claude Opus 5.5

Find out more

Open-Weight Models Can Do Physical AI, but Only With Specialized Harnesses: Can LLMs solve engineering problems effectively without the help of Dyad? Dr. Chris Rackauckas and Venkateshprasad Bhat put leading AI models to the test. They compared how well each model solved real engineering problems on its own and how well it did working inside Dyad. The gap was hard to miss. Dyad turned out to be essential for getting the benefits of AI while keeping the confidence in results that every engineer needs.  Read more

The Perfect Lap: Hunting Down Lost Tenths with Physics-Based Modeling: How fast can a race car get around a track if the driver does everything perfectly? Sebastian Micluța-Câmpeanu and Rajeev Voleti used Dyad to answer that. To check their result, they turned to Dymos, an optimization tool NASA developed for planning things like aircraft and spacecraft trajectories, which includes a well-known race car example that’s become a standard benchmark. Dyad found a faster lap than the NASA reference and solved the problem in a small fraction of the time. Problems that used to take minutes to explore now take seconds, so engineers can test far more design and strategy options. Read more 


Julia 1.13.3 Released and Julia 1.14 Alpha Now Available: Julia 1.13.3 and LTS 1.10.13 are out, and Julia 1.14.0-alpha2 is available for testing. Work by Ian Butterworth and Kristoffer Carlsson suggests Julia 1.14 load and precompile times will be 30–50% lower than 1.10, and Tim Besard's profile-guided optimization (PGO) builds are available through juliaup with the +opt suffix. Julia 1.13.3 | Julia 1.14.0-alpha2

Free Online Julia Webinars from JuliaHub

Events and Conferences

JuliaCon Local Tokyo 2026: JuliaCon Local Tokyo takes place December 4–5, 2026, at the University of Tokyo's Kashiwa Campus, with talks in Japanese and English, tutorials and a poster session. The call for proposals is open until November 15 (23:59 JST), and attendee registration is open now. Find out more

Building Trust in AI for Water Asset Health: Lessons from F1:  At the Water Asset Health Summit 2026, Dr. Criston Hyett explained why the costliest events, like pump failures and burst pipes, are the rarest in the data. His answer: start with known physics, use machine learning to fill the gaps, and show interpretable, uncertainty-aware results. Read more

JuliaCon Global 2026 Videos and Proceedings: JuliaCon Global 2026 talks from Mainz are being released on YouTube, four per day. Presenters can submit papers to the JuliaCon Proceedings until the end of March 2027, and the editorial board is recruiting reviewers. Watch the talks | Proceedings

JuliaCon Global 2027: JuliaCon Global 2027 will take place August 9–14, 2027, at TalTech in Tallinn, Estonia. Find out more

Insights and Recommendations

From Startup to Steady State: HVAC Modeling without Tooling Friction: David Dinh shows how one vapor-compression HVAC model in Dyad handles both transient and steady-state analysis. Switching analysis types changes only the analysis setup, and Dyad Agent generates pressure–enthalpy diagrams with inspectable code. Read more

Replacing PyTorch with Julia and Reactant: One Year In: Caroll Vance describes replacing a PyTorch workflow with a Julia stack built on Reactant and Enzyme. Read more 

Portable Data-Parallel Primitives for the Julia GPU Stack: In his Google Summer of Code 2026 report, Shreyas Hegde adds portable primitives like reverse, findall, accumulate, mapreduce and sorting to AcceleratedKernels.jl 0.5, so GPU backends can share one implementation. Read more

Recent JuliaHub Webinars:  Nearly 100 free one-hour webinars are available, including the recent Apple Silicon GPUs in Julia session with Dr. Tim Besard. Watch here

This Month in Julia World: This Month in Julia World is a newsletter from Stefan Krastanov with up-to-date information about Julia events, new releases and more. Read it here. 

Nouvelles Julia - Julia News en Français: Nouvelles Julia is a newsletter in French with the latest Julia news. Read it here. 

Julia Dispatch Podcast: Hosted by Dr. Chris Rackauckas (JuliaHub VP of Modeling and Simulation) and Dr. Michael Tiemann. Watch here

JuliaHub Digital Twin Solutions and Consulting:  JuliaHub helps enterprises build scalable SciML-based digital twins for asset health, optimization, control and design. Schedule a consultation

Careers at JuliaHub: JuliaHub has fully remote employees across 20 countries. Learn more

Julia Expertise Needed at University of Glasgow: Dr. Eric Silverman seeks a Research Associate for a 5-year agent-based public health modeling project in Julia. A PhD and Julia experience are required. Details and application

Julia and JuliaHub in the News

  • Bioengineer.org: Julia Package Muscade.jl Enables FEM-Constrained Optimisation

  • Quantum Zeitgeist: Maribor Team Builds Quantum Billiards Simulation Toolkit

  • Bioengineer.org: Julia Package Speeds Systemic-Risk Calculations by up to 300-Fold

  • Analytics Insight: Julia Programming Language: How an MIT Project Became a Global Success

  • Slashdot: Programming Language Popularity Increases for R, Julia, and Rust (but Not Go)

  • Quantum Computing Report: IBM Expands Qiskit Beyond Python with Native C API Bindings for Fortran, C++, and Julia

Julia Blog Posts

Contact Us

Contact JuliaHub to learn more about Dyad and JuliaHub, get pricing, schedule training, share case studies, or partner on a Julia event.

About JuliaHub, Julia and Dyad

Dyad combines physics-based modeling with SciML in an agentic engineering tool, and it is free to use. Get started here. JuliaHub is a code-to-cloud platform for Julia. Julia is an open-source, high-performance language for scientific computing, used by more than 10,000 companies and 1,500 universities.

Learn about Dyad: Dyad Studio · Documentation · Community

Learn about Dyad

Get Dyad Studio – Download and install the IDE to start building hardware like software.

Read the Dyad Documentation – Dive into the language, tools, and workflow.

Join the Dyad Community – Connect with fellow engineers, ask questions, and share ideas.

Contact Us

Want to get enterprise support, schedule a demo, or learn about how we can help build a custom solution? We are here to help.