Pick of the Week

Insert Token: 15 Arcade Games in MATLAB with Claude and the MCP Server

Mike Croucher — Wed, 25 Mar 2026 11:56:26 +0000

This week's pick is Arcade by Vasilis Bellos

MATLAB's MCP Server allows AI agents, such as Claude or Amp, to use MATLAB on your machine. In turn, this allows you to develop MATLAB scripts and applications faster than you ever have before. We talk about these new workflows a lot in the Generative AI discussions forum on MATLAB Central. 

MATLAB user Vasilis Bellos has clearly been having a lot of fun with these tools since he has released Arcade: 15 arcade games built entirely in MATLAB - 8 timeless classics and 7 originals. No toolboxes, no external dependencies, no imported assets. Every pixel is drawn using native MATLAB graphics.

Features a neon-styled launcher with an animated starfield menu, persistent high scores, frame-rate-independent physics, and automatic display scaling that ties everything together. Pick up and play with your mouse or keyboard.

I've spent the morning playing these and will be spending the afternoon looking at how Claude and Vasilis got those cool Neon vector effects out of MATLAB Graphics. 

MATLAB user John D’Errico finds largest known reversible prime with 12,346 digits

Mike Croucher — Mon, 16 Mar 2026 15:14:38 +0000

Mike's pick this week is a candidate for the new world-record reversible prime number found by John D'Errico.

A reversible prime, (also known as an emirp), is a prime number that yields a different prime when its digits are reversed. An example is 13. Reversing this gives 31 and both are prime

isprime([13 31])

ans = 1×2 logical array
1 1

Checking if small numbers are emirp is relatively straightforward but things get rather more complex when we start looking at larger numbers. Matt Parker's Numberphile recently discussed the discovery of the (then) largest known emirp. It had 10,002 decimal digits. Wikipedia currently lists 117,954,861 × 1011,111 + 1 as the largest known emirp with which has 11,120 digits.

How MATLAB user, John D'Errico, found a 12,346 digit emirp

The short version of the story is that John looked for emirps of the form 10^n + a and used symbolic toolbox and parallel computing toolbox to make use of all 16 cores on his desktop computer to burn through the search space. The result is that he found that 10^12338*79191501 + 1 is a 12,346 digit emirp, currently the largest known (although I'd prefer us to get independent verification before we suggest an update the Wikipedia page).

Why he chose that particular form of emirp, all of this code and full discussion of his journey can be found in this MATLAB discussion thread. I found it fascinating to follow him on his journey from 'This is a promising candidate' to 'We can say that it's highly probable that this is an emirp' to 'This is definitely an emirp'! Great work John!

We fully expect, and hope, that this record will not stand for long. Happy hunting!

MATLAB on the Steam Deck

Pick of the Week Team — Fri, 06 Mar 2026 15:44:44 +0000

Toshi’s pick this week is MATLAB on the Steam Deck from Reddit by Mindless_Egg7324

A recent Reddit post showed something unexpected and fun: a user successfully installed MATLAB on a Steam Deck, and the community loved it—nearly 500 upvotes and an entire thread of enthusiastic reactions.

The Steam Deck runs SteamOS, which is based on Arch Linux. While MATLAB officially supports a set of validated Linux distributions—including Ubuntu, Debian, Red Hat Enterprise Linux, and SUSE Linux Enterprise Desktop/Server—Arch Linux and SteamOS aren’t on the official support list.

That makes this achievement even more interesting. It works—not because it’s a standard setup, but because someone decided to try it anyway.

This is obviously not practical, but the community had a lot of fun praising the audacity of the OP and cracking jokes on what they might do with it.

The comments captured that feeling:

The excitement came from seeing MATLAB in a completely unexpected place—a gaming handheld, running a non‑standard Linux distro.

It’s a great reminder of how inventive the community can be.

MATLAB Deep Learning Image Classifier App

Mike Croucher — Fri, 27 Feb 2026 17:33:10 +0000

Mike's pick this week is the Deep Learning Image Classifier App by MathWorks Deep Learning Toolbox Team.

Most people who start learning about Deep Learning begin with image classification workflows. I certainly did! My first deep learning experiments involved cats, dogs, digits and tiny little pictures of clothes! 

The MATLAB Deep Learning Image Classifier is a free App Designer app for training image classification deep neural networks in MATLAB. Using this app, you can:

Import, visualize, and augment data
Quickly transfer learn with the SqueezeNet pretrained network
Modify pretrained networks for transfer learning with Deep Network Designer
Import networks from the workspace
Explain predictions with explainability techniques like Grad-CAM and LIME
Generate MATLAB code for training an image classifier

Its a great way to get started with deep learning workflows without worrying about the code; although the code is just a click away when you are ready for it. 

The short video below gives an example of the app in use. In this case, we start off with a pre-trained image classification network called Squeezenet. Squeezenet can classify images into 1000 object categories, such as keyboard, mouse, pencil, and many animals. It cannot, however, classify MathWorks swag!

Using a technique called Transfer Learning, we can adapt squeezenet to recognize the different types of MathWorks swag and the entire workflow can be done in the app.

This just scratches the surface of what this app can do! Since it is all available on GitHub, you can even adapt the app itself to cover workflows that haven't been considered yet.

Agentic AI + MATLAB: A Workflow Worth Exploring

Pick of the Week Team — Sat, 07 Feb 2026 00:09:46 +0000

Toshi’s pick this week is Exploring LLMs in Chemical Engineering: Thermodynamic Modeling meets Vibe Coding from LinkedIn by Antonio Mazzei

Have you ever wondered what happens when you combine agentic AI workflows with MATLAB? A couple months ago, I saw a fantastic example from Antonio Mazzei that shows just how powerful this combination can be.

Antonio used a “vibe coding” approach with Gemini CLI to connect to the MATLAB MCP Core Server and build an interactive script for thermodynamic modeling. The goal was to visualize the 1-parameter Margules model for Excess Gibbs Free Energy—a classic chemical engineering problem.

What’s exciting here isn’t just the math—it’s the workflow. By letting an LLM act as an agent, Antonio didn’t just generate code; he created a system that understands context, iterates on solutions, and integrates with MATLAB. This is a glimpse into the future of technical computing.

The best part? The MATLAB MCP Core is open source and available on GitHub, so you can try this yourself here. Whether you’re in chemical engineering, finance, or any modeling-heavy domain, this approach opens up new possibilities for automation and creativity.

Your turn: Have you experimented with LLMs for modeling or simulation tasks? Share your experience in the comments—I’d love to hear how you’re using AI to push boundaries.

MATLAB DocMaker: making it easier to create documentation for your MATLAB toolbox

Mike Croucher — Fri, 30 Jan 2026 18:57:53 +0000

Mike's pick this week is MATLAB DocMaker by David Sampson. 

Many of the toolboxes featured on Pick Of The Week include superb documentation that integrates with MATLAB's documentation system. They appear in a section called 'Supplemental Software' and here's that section in my MATLAB. You'll recognize a couple of picks from last year such as FSDA Toolbox and k-Wave in the list along with a couple you haven't seen before.

'Koch Toolbox' is a toy I made for myself to see how easy it was to use today's pick: MATLAB DocMaker toolbox. 

It was pretty easy! After installing DocMaker toolbox and connecting it to GitHub (The current version of the toolbox makes use of the GitHub Markdown API), all I needed to do was write a couple of Markdown Files and then ask DocMaker to do its magic. Three commands later and I was done.

DocMaker even uses itself to generate its own documentation, which is also available on GitHub.

With DocMaker, you can:

write documentation in Markdown and convert to HTML for viewing in MATLAB
run MATLAB code blocks in documents and include textual and graphical output
create MATLAB documentation index files from a Markdown table of contents

It also has a cute logo.

DocMaker requires MATLAB R2021a or later to generate documentation but the resulting docs can be viewed offline and in much older MATLAB releases. Check it out on File Exchange and GitHub.

Our first AI generated pick: Double Pendulum Chaos Explorer

Mike Croucher — Fri, 16 Jan 2026 16:57:42 +0000

Mike's pick this week is Double Pendulum Chaos Explorer (AI built) by Duncan Carlsmith.

Today's pick is interesting from at least two points of view; the first of which is the fact that it provides an easy to use application to explore the dynamics of double pendulums. Release two double pendulums from almost, but not quite, identical starting positions and soon they are moving in completely different patterns. With five, things get crazy really fast! It's a classic demonstration of chaos theory. 

The second reason why today's pick is interesting is how it was developed. Duncan has been experimenting with various AI tools such as Claude Desktop and MATLAB MCP Core Server. He has been discussing his experiments over at the MATLAB Central Generative AI discussion forum. The story behind this particular application is at Double Pendulum Chaos Explorer: From HTML5 Prototype to MATLAB interactive application with AI.

2026 is a provably Happy New Year

Mike Croucher — Fri, 09 Jan 2026 17:25:43 +0000

Mike's pick this week is Happy Numbers by Martin Foltin

Every new year, I like to see what makes this year's number interesting. For 2026, I got an easy first hit since it turns out that 2026 is an example of a Happy Number.

According to the Wikipedia page, "A happy number is a number which eventually reaches 1 when the number is replaced by the sum of the square of each digit."  Here's a demonstration that 2026 is happy

2026 →  2² + 0² + 2² + 6² = 44

44 →  4² + 4² = 32

32 →  3² + 2² = 13

13 →  1² + 3² = 10

10 →  1² + 0² = 1

All integers are either happy or unhappy and the next happy new year will be in 2030! For further details, check out the Live Script written by Martin that forms today's pick of the week. You can download it from File Exchange or open it directly in MATLAB Online.

Happy New Year!

Holiday Break with Cody

John Kelly — Thu, 25 Dec 2025 16:57:34 +0000

John's pick this week is to embrace the holiday spirit with three festive Cody problems

Why these picks? As a lot of us break for the holiday season we wanted to combine a coding challenge with some holiday cheer. Whether you’re counting down the days, tallying gifts, or singing along to a classic carol, these problems will keep your MATLAB skills sharp while spreading joy.

Your challenge:
How many of these can you solve? One? All three? Go ahead—grab a warm drink, fire up Cody, and let the holiday coding begin.

Happy Holidays from the Pick of the Week Team

WEC-Sim (Wave Energy Converter Simulator)

Pick of the Week Team — Thu, 11 Dec 2025 12:33:12 +0000

George's pick of the week is WEC-Sim (Wave Energy Converter Simulator) by Sandia National Laboratories and the National Renewable Energy Laboratory (NREL). You can find the source code on GitHub

WEC-Sim (Wave Energy Converter Simulator) is a simulation tool for modelling wave energy converters and offshore systems under realistic wave loading. It is designed specifically for wave-structure interaction and represents devices as interconnected bodies, joints, power-take-off systems, and moorings within a single simulation framework.

Built on MATLAB, Simulink, Simscape, and Simscape Multibody, WEC-Sim lets you simulate a wide variety of real wave-energy concepts: from heaving point absorbers and oscillating water columns (OWCs) to oscillating wave surge converters (OSWECs) and floating devices with realistic mooring systems.

WEC-Sim supports Model Based Design workflows, helping users explore design trade-offs, develop and test control strategies, and estimate power performance in realistic sea states before building hardware. That means faster iteration, lower risk, and better-informed engineering decisions.

The WEC-Sim Applications repository provides a rich set of ready-to-run examples and strong documentation.

If you work anywhere at the intersection of waves, renewables, and power generation, WEC-Sim is a tool you should absolutely have in your modelling toolkit.

Current members of the development team include:

Jeff Grasberger (Sandia), Salman Husain (NREL), Adam Keester (Sandia - PI), Jorge Leon (Sandia), James Niffenegger (NREL), David Ogden (NREL), Mohamed A. Shabara (NREL - PI)

Former members of the development team include:

Dominic Forbush (Sandia), Michael Lawson (NREL), Carlos Michelen (Sandia), Kelley Ruehl (Sandia), Nathan Tom (NREL), Jennifer Van Rij (NREL), Yi-Hsiang Yu (NREL)