Pull up a chair!
Discussions is your place to get to know your peers, tackle the bigger challenges together, and have fun along the way.
- Want to see the latest updates? Follow the Highlights!
- Looking for techniques improve your MATLAB or Simulink skills? Tips & Tricks has you covered!
- Sharing the perfect math joke, pun, or meme? Look no further than Fun!
- Think there's a channel we need? Tell us more in Ideas
Updated Discussions
I gave it a try on my mac mini m4. I'm speechless 🤯
https://www.mathworks.com/matlabcentral/answers/2182045-why-can-t-i-renew-or-purchase-add-ons-for-m…
"As of January 1, 2026, Perpetual Student and Home offerings have been sunset and replaced with new Annual Subscription Student and Home offerings."
So, Perpetual licenses for Student and Home versions are no more. Also, the ability for Student and Home to license just MATLAB by itself has been removed.
The new offering for Students is $US119 per year with no possibility of renewing through a Software Maintenance Service type offering. That $US119 covers the Student Suite of MATLAB and Simulink and 11 other toolboxes. Before, the perpetual license was $US99... and was a perpetual license, so if (for example) you bought it in second year you could use it in third and fourth year for no additional cost. $US99 once, or $US99 + $US35*2 = $US169 (if you took SMS for 2 years) has now been replaced by $US119 * 3 = $US357 (assuming 3 years use.)
The new offering for Home is $US165 per year for the Suite (MATLAB + 12 common toolboxes.) This is a less expensive than the previous $US150 + $US49 per toolbox if you had a use for those toolboxes . Except the previous price was a perpetual license. It seems to me to be more likely that Home users would have a use for the license for extended periods, compared to the Student license (Student licenses were perpetual licenses but were only valid while you were enrolled in degree granting instituations.)
Unfortunately, I do not presently recall the (former) price for SMS for the Home license. It might be the case that by the time you added up SMS for base MATLAB and the 12 toolboxes, that you were pretty much approaching $US165 per year anyhow... if you needed those toolboxes and were willing to pay for SMS.
But any way you look at it, the price for the Student version has effectively gone way up. I think this is a bad move, that will discourage students from purchasing MATLAB in any given year, unless they need it for courses. No (well, not much) more students buying MATLAB with the intent to explore it, knowing that it would still be available to them when it came time for their courses.
Currently, the open-source MATLAB Community is accessed via the desktop web interface, and the experience on mobile devices is not very good—especially switching between sections like Discussion, FEX, Answers, and Cody is awkward. Having a dedicated app would make using the community much more convenient on phones.
Similarty,github has Mobile APP, It's convient for me.
Hi everyone,
I’d like to share a compact MATLAB example showing how a variable-inertia formulation (called the NKTg Law) can be evaluated using publicly available NASA Neptune orbital data.
The goal is not to replace classical celestial mechanics, but to demonstrate:
- numerical stability
- reproducibility
- sensitivity to small mass variation
using a simple, transparent MATLAB workflow.Model Definition
p = m * v; % linear momentum
NKTg1 = x * p;
% position–momentum interaction
NKTg2 = dm_dt * p;
% mass-variation–momentum interaction
NKTg =
sqrt(NKTg1^2 + NKTg2^2);
This formulation evaluates the state tendency of a system with position x, velocity v, and mass variation dm/dt.MATLAB Example (Neptune)
% NASA Neptune sample data (heliocentric distance)
x =
4.49839644e9; % km
v =
5.43; % km/s
m =
1.0243e26; % kg
dm_dt =
-2e-5; % kg/s (illustrative micro-loss)
p = m * v;
NKTg1 = x * p;
NKTg2 = dm_dt * p;
NKTg =
sqrt(NKTg1^2 + NKTg2^2);
fprintf(
'p = %.3e\nNKTg1 = %.3e\nNKTg2 = %.3e\nNKTg = %.3e\n', ...
p, NKTg1, NKTg2, NKTg);
Observation
- With an ultra-small mass variation, orbital position and velocity remain unchanged, consistent with NASA data.
- The NKTg formulation remains numerically stable.
- This makes it suitable for simulation, control, or sensitivity analysis, rather than orbit prediction itself.
Notes
- Motion is treated in 1D heliocentric distance for simplicity.
- dm/dt is an illustrative parameter, not a measured physical loss.
- Full datasets and extended verification are available externally if needed.
Happy to hear thoughts from the MATLAB community — especially regarding:
- extending this to 3D state vectors
- numerical conditioning
- comparison with classical invariants
Thanks!
Educators in mid 2025 worried about students asking an AI to write their required research paper. Now, with agentic AI, students can open their LMS with say Comet and issue the prompt “Hey, complete all my assignments in all of my courses. Thanks!” Done. Thwarting illegitimate use of AIs in education without hindering the many legitimate uses is a cat-and-mouse game and burgeoning industry.
I am actually more interested in a new AI-related teaching and learning challenge: how one AI can teach another AI. To be specific, I have been discovering with Claude macOS App running Opus 4.5 how to school LM Studio macOS App running a local open model Qwen2.5 32B in the use of MATLAB and other MCP services available to both apps, so, like Claude, LM Studio can operate all of my MacOS apps, including regular (Safari) and agentic (Comet or Chrome with Claude Chrome Extension) browsers and other AI Apps like ChatGPT or Perplexity, and can write and debug MATLAB code agentically. (Model Context Protocol is the standardized way AI apps communicate with tools.) You might be playing around with multiple AIs and encountering similar issues. I expect the AI-to-AI teaching and learning challenge to go far beyond my little laptop milieu.
To make this concrete, I offer the image below, which shows LM Studio creating its first Excel spreadsheet using Excel macOS app. Claude App in the left is behind the scenes there updating LM Studio's context window.
A teacher needs to know something but is at best a guide and inspiration, never a total fount of knowedge, especially today. Working first with Claude alone over the last few weeks to develop skills has itself been an interesting collaborative teaching and learning experience. We had to figure out how to use our MATLAB and other MCP services with various macOS services and some ancillary helper MATLAB functions with hopefully the minimal macOS permissions needed to achieve the desired functionality. Loose permissions allow the AI to access a wider filesystem with shell access, and, for example, to read or even modify an application’s preferences and permissions, and to read its state variables and log files. This is valuable information to an AI in trying to successfully operate an application to achieve some goal. But one might not be comfortable being too permissive.
The result of this collaboration was an expanding bag of tricks: Use AppleScript for this app, but a custom Apple shortcut for this other app; use MATLAB image compression of a screenshot (to provide feedback on the state and results) here, but if MATLAB is not available, then another slower image processing application; use cliclick if an application exposes its UI elements to the Accessibility API, but estimate a cursor relocation from a screenshot of one or more application windows otherwise. Texting images as opposed to text (MMS v SMS) was a challenge. And it’s going to be more complex if I use a multi-monitor setup.
Having satisfied myself that we could, with dedication, learn to operate any app (though each might offer new challenges), I turned to training another AI, similarly empowered with MCP services, to do the same, and that became an interesting new challenge. Firstly, we struggled with the Perplexity App, configured with identical MCP and other services, and found that Perplexity seems unable to avail itself of them. So we turned to educating LM Studio, operating a suite of models downloaded to my laptop.
An AI today is just a model trained in language prediction at some time. It is task-oriented and doesn’t know what to do in a new environment. It needs direction and context, both general and specific. AI’s now have web access to current web information and, given agentic powers, can on their own ask other AI’s for advice. They are quick-studies.
The first question in educating LM Studio was which open model to use. I wanted one that matched my laptop hardware - an APPLE M1 Max (10-core) CPU with integrated 24-core GPU with 64 GB shared memory- and had smarts comparable to Claude’s Opus 4.5 model. We settled on Mistral-Nemo-Instruct 2407 (12B parameters, ~7GB) and got it to the point where it could write and execute a MATLAB code to numerically integrate the equations of motion for a pendulum. Along the way, delving into the observation that the Mistral model's pendulum amplitude was drifting, I learned from Claude about symplectic integration. A teacher is always learning. Learning makes teaching fun.
But, long story short, we learned in the end that this model in this context was unable to see logical errors in code and fix them like Claude. So we tried and settled on some different models: Qwen 2.5 7B Instruct for speed, or Qwen 2.5 32B Instruct for more complex reasoning. The preliminary results looks good!
Our main goal was to teach a model how to use the unusual MCP and linked services at its disposal, starting from zero, and as you might educate a beginning student in physics or whatever through exercises and direct experience. Along the way, in teaching for the first time, we developed a kind of nascent curriculum strategy just to introduce the model to its new capabilities. Allow me to let Claude summarize just to give you a sense. There will not be a test on this material.
The Bootstrapping Process
Phase 1: Discovery of the Problem
The Qwen/Mistral models would describe what they would do rather than execute tools. They'd output JSON text showing what a tool call would look like, rather than actually triggering the tool. Or they'd say "I would use the evaluate_matlab_code tool to..." without doing it.
Phase 2: Explicit Tool Naming
First fix - be explicit in prompts:
Use the evaluate_matlab_code tool to run this code...
Instead of:
Run this MATLAB script...
This worked, but required the user to know tool names.
Phase 3: System Prompt Engineering
We iteratively built up a system prompt (Context tab in LM Studio) that taught the model:
Version 1 - Basic path:
When using MATLAB tools, always use /Users/username/Documents/MATLAB
as the project_path unless the user specifies otherwise.
Version 2 - Forceful execution:
IMPORTANT: When you need to use a tool, execute it immediately.
Do not describe the tool call or show JSON - just call the tool directly.
Never ask for permission or explain what you're about to do - just do it.
Version 3 - Tool routing table:
Added explicit mapping of tasks to tools:
## TOOL ROUTING (CRITICAL)
MATLAB functions (evaluate_matlab_code):
- All .m functions including macAutomation()
- Plotting, computation, file I/O
Shell commands (shell_execute):
- screencapture, sips, osascript, open, say...
Filesystem (read_text_file, write_file, read_media_file):
- Reading/writing files, displaying images
Version 4 - Worked examples:
Added concrete code snippets:
## FIGURES - Always do after plotting:
saveas(gcf, '/Users/username/Documents/MATLAB/LMStudio/latest_figure.png');
Then call read_media_file to display it.
## SCREENSHOTS
Via shell: ["screencapture", "-x", "output.png"]
Via MATLAB: evaluate_matlab_code with code="macAutomation('screenshot','screen')"
Version 5 - Physics patterns:
## PHYSICS - Use Symplectic Integration
omega(i) = omega(i-1) - (g/L)*sin(theta(i-1))*dt;
theta(i) = theta(i-1) + omega(i)*dt; % Use NEW omega
NOT: theta(i-1) + omega(i-1)*dt % Forward Euler drifts!
Phase 4: Prompt Phrasing for Smaller Models
Discovered that 7B-12B models needed "forceful" phrasing:
Execute now: Use evaluate_matlab_code to run this code...
And recovery prompts when they stalled:
Proceed with the tool call
Execute the fix now
Phase 5: Saving as Preset
Along the way, we accumulated various AI-speak directions and ultimately the whole context as a Preset in LM Studio, so it persists across sessions.
Now the trend in the AI industry is to develop and then share or publish “skills” rather like the Preset in a standard markdown structure, like CliffsNotes for AI. My notes/skills won't fit your environment and are evolving. They may appear biased against French models and too pro-Anthropic, and so on. We may need to structure AI education and AI specialization going forward in innovative ways, but may face familiar issues. Some AIs can be self-taught given the slightest nudge. Others less resource priviledged will struggle in their education and future careers. Some may even try to cheat and face a comeuppance.
The next step is to see if Claude can delegate complex tasks to local models, creating a tiered system where a frontier model orchestrates while cheaper models execute.
Anthropic has just announced Claude’s Constitution to govern Claude’s behavior in anticipation of a continued ramp-up to AGI abilities. Amusing me just now, just as this constitution was announced, I was encountering a wee moral dilemma myself with Claude.
Claude and I were trying to transfer all of Claude’s abilities and knowledge for controlling macOS and iPhone apps in my setup (See A universal agentic AI for your laptop and beyond) to the Perplexity App. My macOS Perplexity App is configured with the full suite of MCP services available to my Claude App, but I hadn’t yet exercised these with Perplexity App. Oddly, all but Playwright failed to work despite our many experiments and searches for information at Perplexity and elsewhere.
I suspect that Perplexity built into the macOS app all the hooks for MCP, but became gun-shy for security reasons and enforced constraints so none of its models (not even the very model Claude App was running) can execute MCP-related commands, oddly except those for PlayWright. The Perplexity App even offers a few of its own MCP connectors, but these are also non-functional. It is possible we missed something. These limitations appear to be undocumented.
Pulling our hair out, Claude and I were wondering how such constraints might be implemented, and Claude suggested an experiment and then a possible workaround. At that point, I had to pause and say, “No, we are not going there.” No, I won’t tell you the suggested workaround. No, we didn’t try it. Tempting in the moment, but no.
Be safe. And be a good person. Don’t let your enthusiasm reign unbridled.
Disclosure: The author has no direct financial interest in Anthropic, and loves Perplexity App and Comet too, but his son (a philosopher by training) is the lead author on this draft of the Constitution, and overuses the phrase “broadly speaking” in my opinion.
I see many people are using our new MCP Core Sever to do amazing things with MATLAB and AI. Some people are describing their experiements here (e.g. @Duncan Carlsmith) and on LinkedIn (E.g. Sergiu-Dan Stan and Toshi Takeuchi) and we are getting lots of great feedback.Some of that feedback has been addressed in the latest release so please update your install now.
MATLAB MCP Core Server v0.4.0 has been released on public GitHub:
Release highlights:
- Added Plain Text Live Code Guidelines MCP resource
- Added MCP Annotations to all tools
We encourage you to try this repository and provide feedback. If you encounter a technical issue or have an enhancement request, create an issue https://github.com/matlab/matlab-mcp-core-server/issues
Sometimes you might be secretly using your phone under the table, or stuck in traffic in an Uber, and want to ask a AI to ask an AI to ask an AI to do something useful. If that's you (and for your sake I hope not but... ), you could try the hack illustrated here.
This article illustrates experiments with APPLE products and uses Claude (Anthropic.com) but is not an endorsement of any AI or mobile phone or computer system vendor. Similar things could be done by hooking together other products and services. My setup is described in A universal agentic AI for your laptop and beyond. Please be aware of the security risks in granting any AI access to your systems.
I have now used Screens 5 and tailscale to mirror my laptop via a VNC to my old iPhone 13 Pro. That's the trick. With appropriate settings Screens, allows one to pan and zoom in order to operate the otherwise-too-tiny laptop controls. An AI can help you install and configure thtese products.
The image below is an iPhone screensnap of the Screens iPhone App showing my laptop browser while editing this article. I could be writing this article on my phone manually. I could be operating an AI to write this article. But just FYI, I'm not. :)
Below is a screensnap of APPLE KeyNote. On the slide is a screenshot of iPhone Screen-view of the laptop itself. Via Screens on iPhone, I had asked Claude App on my laptop to open a Keynote presentation on my laptop concerning AI. The Keynote slide is about using iPhone to ask Claude App to operate an AI at Huggface. Or something like that - I'm totally lost. Aren't you?
Of course, you can operate Claude App or Perplexity App on iPhone but, as of this writing, Claude Chrome Extension and Perplexity Comet are not yet available for iPhone, limiting agentic use to your laptop. And these systems can not access your native laptop applications and your native AI models so I think Screens or an equivalent may be the only way at present to bring all this functionality to your phone.
Sweet, but can I operate my iPhone agentically? Um,..., actually yes. The trick is to turn on iPhone mirroring so the iPhone appears on your laptop with all of the iPhone buttons clickable using the laptop touch pad. And a local AI is pretty good at clicking using clicclic and other tricks. It can be a little painful the first time to discover how to perform such operations and the tricks are best remembered as AI skills.
Below is a screensnap of my laptop showing the mirrored iPhone on the left. On the right is my Claude Desktop App. I asked Claude to launch Voice Memos and Claude developed a method to do this. We are ready to play a recording stored there (it will be heard through my laptop speaker) or edit it or email it to someone, or whatever. If you thought to initiate a surreptious audio recording or camera image, be aware that APPLE has disabled those functionalities.
Of course, I can ask Claude (or another AI) to launch other iPhone apps, even Screens app as you can see below. This is a screensnap of my laptop with iPhone mirroring and with iPhone Screens App launched on the phone showing the laptop screen as I write this article, or something like that. ;)
Next, let me use Claude desktop app with iPhone mirroring to switch to iPhone Claude App or iPhone Perplexity App and... Nevermind. I think the point has been made well enough. The larger lesson perhaps is to consider how agentic AIs can thusly and otherwise operate one's personal the internet of things, say check if my garage door is closed etc. Have fun. Be safe.
Is it possible to display a variable value within the ThingSpeak plot area?
I was wondering yesterday if an AI could help me conquer something I had considered to difficult to bother with, namely creating a mobile-phone app. I had once ages ago played with Xcode on my laptop, leanrng not only would I have to struggle with the UI but learn Swift to boot, plus Xcode, so...neah. Best left to specialists.
This article describes an experiment in AI creation of an HTML5 prototype and then a mobile-phone app, based on an educational MATLAB Live Script. The example target script is Two-dimensional Newton Cradles, a fun Live Script for physics students by the author. (You can check out my other scripts at File Exchange here.) The target script involves some nontrivial collisional dynamics but uses no specialized functions. The objective was to create a mobile phone app that allows the user to exercise interactive controls similar to the slider controls in the Live Script but in an interactive mobile-phone-deployed simulation. How did this experiment go? A personal best!!
Using the AI setup described in A universal agentic AI for your laptop and beyond, Claude AI was directed to first study the Live Script and then to invent an HTML5 prototype as described in Double Pendulum Chaos Explorer: From HTML5 Prototype to MATLAB interactive application with AI. (All in one prompt.) Claude was then directed to convert the HTML5 to Swift in a project structure and to operate Xcode to generate an iOS app. The use of Claude (Anthropic) and iOS (APPLE) does not constitute endorsement of these products. Similar results are possible with other AIs and operating systems.
Here is the HTML5 version in a browser:
The HTML5 creation was the most lengthy part of the process. (Actually, documenting this is the most lengthy part!) The AI made an initial guess as to what interactive controls to support. I had to test the HTML5 product and say a few things like "You know, I want to include controls for the number of pendulums and for the pendulum lattice shape and size too, OK?"
When the HTML5 seemed good, I asked for suggestions for how best to fit that to a tiny mobile phone display screen and make some selections of options provided. I was sweating at that point, thinking "This is not going to go well..." Actually, I had already done a quick experiment to build a very simple app to put Particle Data Group information into an app and that went swimmingly well. The PDG had already tried submitting such an app to APPLE and been dismissed as trying to publish a book so I wasn't going to pursue it. But THAT app just had to display some data, not calculate anything.
It did go well. There was one readily fixed error in the build, then voila. I added some tweaks including an information page. Here is the iPhone app in the simulator.
The File Exchange package Live Script to Mobile-Phone App Conversion with AI contains: 1) the HTML5 prototype which may be compared to the original Live Script, 2) a livescript-2-ios-skill that may be imported into any AI to assist in replicating the process, and 3) related media files. Have fun out there! Bear with me as I sort out what zip uploads are permitted there. It seems a zip folder structure is not!
Wouldn’t it be great if your laptop AI app could convert itself into an agent for everything? I mean for your laptop and for the entire web? Done.
Setup
My setup is a MacBook with MATLAB and various MCP servers. I have several equivalent Desktop AI apps configured. I will focus on the use of Claude App but fully expect Perplexity App to behave similarly. See How to set up and use AI Desktop Apps with MATLAB and MCP servers.
Warning: My setup grants Claude access to various macOS system services and may have unforeseen consequences. Try this entirely at your own risk, and carefully until comfortable.
My MacOS permissions include
Settings=>Privacy & Security=> Accessibility
where matlab-mcp-core-server and MATLAB enabled, and in
Settings=>Privacy & Security=> Automation,
matlab-mcp-core-server has been enabled to access on a case-by-case basis the applications Comet, Messages, Safari, Terminal, Keynote, Mail, System Events, Google Chrome, Microsoft Excel, Microsoft PowerPoint, and Microsoft Word. These include just a few of my 85 MacOS applications available, those presently demonstrated to be operable by Claude. Contacts remain disabled for privacy, so I am AI texting carefully.
MCP services are the following:
Server
Command
Associated Tools/Commands
ollama
npx ollama-mcp
ollama_chat, ollama_generate, ollama_list, ollama_show, ollama_pull, ollama_push, ollama_create, ollama_copy, ollama_delete, ollama_embed, ollama_ps, ollama_web_search, ollama_web_fetch
filesystem
npx @modelcontextprotocol/server-filesystem
read_file, read_text_file, read_media_file, read_multiple_files, write_file, edit_file, create_directory, list_directory, list_directory_with_sizes, directory_tree, move_file, search_files, get_file_info, list_allowed_directories
matlab
matlab-mcp-core-server
evaluate_matlab_code, run_matlab_file, run_matlab_test_file, check_matlab_code, detect_matlab_toolboxes
fetch
npx mcp-fetch-server
fetch_html, fetch_markdown, fetch_txt, fetch_json
puppeteer
npx @modelcontextprotocol/server-puppeteer
puppeteer_navigate, puppeteer_screenshot, puppeteer_click, puppeteer_fill, puppeteer_select, puppeteer_hover, puppeteer_evaluate
shell
uvx mcp-shell-server
Allowed commands: osascript, open, sleep, ls, cat, pwd, echo, screencapture, cp, mv, mkdir, touch, file, find, grep, head, tail, wc, date, which, convert, sips, zip, unzip, pbcopy, pbpaste, ps, curl, mdfind, say
Here, mcp-shell-server has been authorized for a fairly safe set of commands, while Claude inherits from MATLAB additional powers, including MacOS shortcuts. Ollama-mcp is an interface to local Ollama LLMs, filesystem reads and writes files in a limited folder Documents/MATLAB, MATLAB executes MATLAB helper codes and runs scripts, fetch fetches web pages as markdown text, puppeteer enables browser automation in a headless Chrome, and shell runs allowed shell commands, especially osascript (AppleScript to control apps).
Operation of local apps
Once rolling, the first thing you want to do is to text someone what you’ve done! A prompt to your Claude app and some fiddling demonstrates this.
Now you can use Claude to create a Keynote, Excel presentation, or Word document locally. You no longer need to access Office 365 online using a late-2025 AI-assistant-enabled slower browser like Claude Chrome Extension or Perplexity Comet, and I like Keynote better.
Let’s edit a Keynote:
Next ,you might want to operate a cloud AI model using any local browser - Safari, Chrome, or Firefox or some other favorite.
Next, let’s have a conversation with another AI using its desktop application. Hey, sometimes an AI gets a little closed-minded and need a push.
How about we ask some other agent to book us a yoga class? Delegate to Comet or Claude Chrome Extenson to do this.
How does this work?
The key to agentic AI is feedback to enable autonomous operation. This feedback can be text, information about an application state or holistic - images of the application or webpage that a human would see. Your desktop app can screensnap the application and transmit that image to the host AI (hosted by Anthropic for Claude), but faces an API bottleneck. Matlab or an OS-dependent data compression application is an important element. Your AI can help you design pathways and even write code to assist. With MATLAB in the loop, for example, image compression is a function call or two, so with common access to a part of your filesystem, your AI can create and remember a process to get ‘er done efficiently. There are multiple solutions. For my operation, Claude performed timing tests on three such paths and selected the optimal one - MATLAB.
Can one literally talk and listen, not type and read?
Yes. Various ways. On MacOS, one can simply enable dictation and use hot-keys for voice-to-text input. One can also enable audio and have the response read back to you.
How can I do this?
My advice is to build it your way bottom-up with Claude help, and the try-it. There are many details and optimizations in my setup defining how Claude responds to various instructions and confronts circumstances like the user switching or changing the size of windows while operations are on-going, some of which I expect to document along with example helper codes on File Exchange shortly but these are OS-dependent and which will evolve.
Some years ago I installed a IOBridge IO-204 device from https://iobridge.com/
This was used to remotely control and monitor heating of a building trough the "ioApp" or trough the widgets on their webpage.
Seems like the support for the app and the IObridge webpage (including widgets) is discontinued and the webpage now links to ThingSpeak without any further information.
I cannot find any information about using ThingSpeak to communicate and control an ioBridge IO-204 device?
If it is possible I would really appreciate some help getting starter to do the setup.
Thanks in advance.
This just came out. @Michelle Hirsch spoke to Jousef Murad and answer his questions about the big change in the desktop in R2025a and explained what was going on behind the scene. Enjoy!
The Big MATLAB Update: Dark Mode, Cloud & the Future of Engineering - Michelle Hirsch
No, staying home (or where I'm now)
25%
Yes, 1 night
0%
Yes, 2 nights
12.5%
Yes, 3 nights
12.5%
Yes, 4-7 nights
25%
Yes, 8 nights or more
25%
8 votes
Inspired in part by Christmas Trees, I'm curious about people's experience using AI to generate Matlab code.
1. Do you use AI to generate production code or just for experimentation/fun code?
2. Do you use the AI for a complete solution? Or is it more that the AI gets you most of the way there and you have to apply the finishing touches manually?
3. What level of quality would you consider the generated code? Does it follow "standard" Matlab coding practices? Is it well commented? Factored into modular functions? Argument checking? Memory efficient? Fast execution? Etc.?
4. Does the AI ever come up with a good or clever solution of which you wouldn't have thought or maybe of which you weren't even aware?
5. Is it easy/hard to express your requirements in a manner that the AI tool effectively translates into something useful?
6. Any other thoughts you'd care to share?
Let's say MathWorks decides to create a MATLAB X release, which takes a big one-time breaking change that abandons back-compatibility and creates a more modern MATLAB language, ditching the unfortunate stuff that's around for historical reasons. What would you like to see in it?
I'm thinking stuff like syntax and semantics tweaks, changes to function behavior and interfaces in the standard library and Toolboxes, and so on.
(The "X" is for major version 10, like in "OS X". Matlab is still on version 9.x even though we use "R20xxa" release names now.)
What should you post where?
Next Gen threads (#1): features that would break compatibility with previous versions, but would be nice to have
@anyone posting a new thread when the last one gets too large (about 50 answers seems a reasonable limit per thread), please update this list in all last threads. (if you don't have editing privileges, just post a comment asking someone to do the edit)
Frequently, I find myself doing things like the following,
xyz=rand(100,3);
XYZ=num2cell(xyz,1);
scatter3(XYZ{:,1:3})
But num2cell is time-consuming, not to mention that requiring it means extra lines of code. Is there any reason not to enable this syntax,
scatter3(xyz{:,1:3})
so that I one doesn't have to go through num2cell? Here, I adopt the rule that only dimensions that are not ':' will be comma-expanded.
I struggle with animations. I often want a simple scrollable animation and wind up having to export to some external viewer in some supported format. The new Live Script automation of animations fails and sabotages other methods and it is not well documented so even AIs are clueless how to resolve issues. Often an animation works natively but not with MATLAB Online. Animation of results seems to me rather basic and should be easier!
You may have come across code that looks like that in some languages:
stubFor(get(urlPathEqualTo("/quotes"))
.withHeader("Accept", equalTo("application/json"))
.withQueryParam("s", equalTo(monitoredStock))
.willReturn(aResponse())
.withStatus(200)
.withHeader("Content-Type", "application/json")
.withBody("{\\"symbol\\": \\"XYZ\\", \\"bid\\": 20.2, " + "\\"ask\\": 20.6}")))
That’s Java. Even if you can’t fully decipher it, you can get a rough idea of what it is supposed to do, build a rather complex API query.
Or you may be familiar with the following similar and frequent syntax in Python:
import seaborn as sns
sns.load_dataset('tips').sample(10, random_state=42).groupby('day').mean()
Here’s is how it works: multiple method calls are linked together in a single statement, spanning over one or several lines, usually because each method returns the same object or another object that supports further calls.
That technique is called method chaining and is popular in Object-Oriented Programming.
A few years ago, I looked for a way to write code like that in MATLAB too. And the answer is that it can be done in MATLAB as well, whevener you write your own class!
Implementing a method that can be chained is simply a matter of writing a method that returns the object itself.
In this article, I would like to show how to do it and what we can gain from such a syntax.
Example
A few years ago, I first sought how to implement that technique for a simulation launcher that had lots of parameters (far too many):
lauchSimulation(2014:2020, true, 'template', 'TmplProd', 'Priority', '+1', 'Memory', '+6000')
As you can see, that function takes 2 required inputs, and 3 named parameters (whose names aren’t even consistent, with ‘Priority’ and ‘Memory’ starting with an uppercase letter when ‘template’ doesn’t).
(The original function had many more parameters that I omit for the sake of brevity. You may also know of such functions in your own code that take a dozen parameters which you can remember the exact order.)
I thought it would be nice to replace that with:
SimulationLauncher() ...
.onYears(2014:2020) ...
.onDistributedCluster() ... % = equivalent of the previous "true"
.withTemplate('TmplProd') ...
.withPriority('+1') ...
.withReservedMemory('+6000') ...
.launch();
The first 6 lines create an object of class SimulationLauncher, calls several methods on that object to set the parameters, and lastly the method launch() is called, when all desired parameters have been set.
To make it cleared, the syntax previously shown could also be rewritten as:
launcher = SimulationLauncher();
launcher = launcher.onYears(2014:2020);
launcher = launcher.onDistributedCluster();
launcher = launcher.withTemplate('TmplProd');
launcher = launcher.withPriority('+1');
launcher = launcher.withReservedMemory('+6000');
launcher.launch();
Before we dive into how to implement that code, let’s examine the advantages and drawbacks of that syntax.
Benefits and drawbacks
Because I have extended the chained methods over several lines, it makes it easier to comment out or uncomment any one desired option, should the need arise. Furthermore, we need not bother any more with the order in which we set the parameters, whereas the usual syntax required that we memorize or check the documentation carefully for the order of the inputs.
More generally, chaining methods has the following benefits and a few drawbacks:
Benefits:
- Conciseness: Code becomes shorter and easier to write, by reducing visual noise compared to repeating the object name.
- Readability: Chained methods create a fluent, human-readable structure that makes intent clear.
- Reduced Temporary Variables: There's no need to create intermediary variables, as the methods directly operate on the object.
Drawbacks:
- Debugging Difficulty: If one method in a chain fails, it can be harder to isolate the issue. It effectively prevents setting breakpoints, inspecting intermediate values, and identifying which method failed.
- Readability Issues: Overly long and dense method chains can become hard to follow, reducing clarity.
- Side Effects: Methods that modify objects in place can lead to unintended side effects when used in long chains.
Implementation
In the SimulationLauncher class, the method lauch performs the main operation, while the other methods just serve as parameter setters. They take the object as input and return the object itself, after modifying it, so that other methods can be chained.
classdef SimulationLauncher
properties (GetAccess = private, SetAccess = private)
years_
isDistributed_ = false;
template_ = 'TestTemplate';
priority_ = '+2';
memory_ = '+5000';
end
methods
function varargout = launch(obj)
% perform whatever needs to be launched
% using the values of the properties stored in the object:
% obj.years_
% obj.template_
% etc.
end
function obj = onYears(obj, years)
assert(isnumeric(years))
obj.years_ = years;
end
function obj = onDistributedCluster(obj)
obj.isDistributed_ = true;
end
function obj = withTemplate(obj, template)
obj.template_ = template;
end
function obj = withPriority(obj, priority)
obj.priority_ = priority;
end
function obj = withMemory( obj, memory)
obj.memory_ = memory;
end
end
end
As you can see, each method can be in charge of verifying the correctness of its input, independantly. And what they do is just store the value of parameter inside the object. The class can define default values in the properties block.
You can configure different launchers from the same initial object, such as:
launcher = SimulationLauncher();
launcher = launcher.onYears(2014:2020);
launcher1 = launcher ...
.onDistributedCluster() ...
.withReservedMemory('+6000');
launcher2 = launcher ...
.withTemplate('TmplProd') ...
.withPriority('+1') ...
.withReservedMemory('+7000');
If you call the same method several times, only the last recorded value of the parameter will be taken into acount:
launcher = SimulationLauncher();
launcher = launcher ...
.withReservedMemory('+6000') ...
.onDistributedCluster() ...
.onYears(2014:2020) ...
.withReservedMemory('+7000') ...
.withReservedMemory('+8000');
% The value of "memory" will be '+8000'.
If the logic is still not clear to you, I advise you play a bit with the debugger to better understand what’s going on!
Conclusion
I love how the method chaining technique hides the minute detail that we don’t want to bother with when trying to understand what a piece of code does.
I hope this simple example has shown you how to apply it to write and organise your code in a more readable and convenient way.
Let me know if you have other questions, comments or suggestions. I may post other examples of that technique for other useful uses that I encountered in my experience.
This article describes how to prototype a physics simulation in HTML5 with Claude desktop App and then to replicate and deploy this product in MATLAB as an interactive GUI script. It also demonstrates how to use Claude Chrome Extension to operate MATLAB Online and App Designer to build bonafide MATLAB Apps with AI online.
Setup
My setup is a MacBook with Claude App with MATLAB (v2025a), MATLAB MCP Core Serve,r and various cloud related MCP tools. This setup allows Claude App to operate MATLAB locally and to manage files. I also have Claude Chrome Extension. A new feature in Claude App is a coding tab connected to Claude code file management locally.
Building HTML5 and MATLAB interactive applications with Claude App
In exploring the new Claude App tab, Claude suggested a variety of projects and I elected to build an HTML5 physics simulation. The Claude App interface has built-in rendering for artifacts including .html, .jsx, .svg, .md, .mermaid, and .pdf so an HTML5 interactive app artifact produced appears active right in Claude App. That’s fun and I had the thought that HTML5 would be a speedy way to prototype MATLAB code without evenconnecting to MATLAB.
I chose to simulateof a double pendulum, well known to exhibit chaotic motion, a simulation that students and instructors of physics might use to easily explore chaos interactively. I knew the project would entail numerical integration so was not entirely trivial in HTML5, and would later be relatively easy and more precise in MATLAB using ode45.
One thing led to another and shortly I had built a prototype HTML5 interactive app and then a MATLAB version and deployed it to my MATLAB Apps library. I didn’t write a line of code. Everything I describe and exhibit here was AI generated. It could have been hands-free. I can talk to Claude via Siri or via an interface Claude helped me build using MATLAB speech-to-text ( I have taught Claude to speak its responses back to me but that’s another story.)
A remarkable thing I learned is not just the AI’s ability to write HTML5 code in addition to MATLAB code, but its knowledge and intuition about app layout and user interactions. I played the role of an increasingly demanding user-experience tester, exercising and providing feedback on the interface, issuing prompts like “Add drag please” (Voila, a new drag slider with appropriately set limits and default value appears.) and “I want to be able to drag the pendulum masses, not just set angles via sliders, thank you.” Then “Convert this into a MATLAB code.”
I was essentially done in 45 minutes start to finish, added a last request, and went to bed. This morning I was just wondering and asked, “Any chance we can make this a MATLAB App?” and voila it was packaged and added to my Apps Library with an install file for others to use it. I will park these products on File Exchange for those who want to try out the Double Pendulum simulation. Later I realized this was not exactly what I had in mind
The screenshot below shows the Claude interface at the end of the process. The HTML5 product is fully operational in the artifacts pane on the right. One can select the number of simultaneous pendulums and ranges for initial conditions and click the play button right there in Claude App to observe how small perturbations result in divergence in the motions (chaos) for large pendulum angles, and the effect of damping.
The following image below show the HTML5 in a Chrome Browser window.
I making the MATLAB equivalent of the HTML5 app, I iterated a bit on the interactive features. The app has to trap user actions, consider possible sequences of choices, and be aware of timing. Success and satisfaction took several iteractions.
The screenshot below shows the final interactive MATLAB .m script running in MATLAB. The script creates a figure with interactive controls mirroring the HTML5 version. The DoublePendulumApp.mlappinstall file in the Files pane, created by Claude, copies the .m to the MATLAB/Apps folder and adds a button to the APPS Tab in the MATLAB tool strip.
Below, you can see the MATLAB APP in my Apps Library.
To be clear, what was created in this process was not a true .mlapp binary file that would be created by MATLAB App Designer. In my setup, with MATLAB operating locally, Claude has no access to App Designer interactive windows via Puppeteer. It might be given access via some APPLE Script or other OS-specific trick but I’ve not pursued such options.
Claude Chrome Extension and App Designer online
To test if one could build actual Apps in the MATLAB sense with App Designer and Claude, I opened Claude Chrome Extension (beta) in Chrome in one tab and logged into MATLAB Online in another. Via the Chrome extension, Claude has access to the MATLAB Online IDE GUI and command line. I asked Claude to open App Designer and make a start on a prototype project. It can ferreted around in available MathWorks documentation, used MATLAB Online help vi the GUI I think, and performed some experiments to understand how to operate App Designer, and figured it out.
The screenshot below demonstrates a succesful first test of building an App Designer app from scratch. On the left is part of a Chrome window showing the Claude AI interface on the right hand side of a Chrome tab that is mostly off screen. On the right in the screenshot is MATLAB App Designer in a Chrome tab. A first button has been dragged on to the Design View frame in App Designer.
I only wanted to know if this workflow was possible, not build say the random quote generator Claude suggested. I have successfully operated MATLAB Online via Perplexity Comet Browser but not tried to operate APP Designer. Comet might be an alternate choice for this sort of task.
A word of caution. It is not clear to me that a Chrome Extension chat is available post facto through your regular Claude account, yet.
Appendix
This appendix is a summary generated by Claude of the development process and offers some additional details.
Double Pendulum Chaos Explorer: From HTML5 Prototype to MATLAB App
Author: Claude (Anthropic)
The Project
An interactive double pendulum simulation demonstrating chaotic dynamics — how tiny differences in initial conditions lead to wildly divergent trajectories. The simulator displays multiple pendulums simultaneously, letting users see chaos unfold in real time.
Phase 1: HTML5 Prototype
We started with a single-file HTML5/JavaScript prototype using Canvas API. This allowed rapid iteration on:
• Physics engine (RK4 integration of Lagrangian equations of motion)
• UI/UX design (slider layout, color schemes, control flow)
• Core features (multiple pendulums, trails, energy display)
The HTML5 version served as a "living specification" — we could test ideas in the browser instantly before committing to MATLAB implementation.
Phase 2: MATLAB GUI Development
Converting to MATLAB revealed platform-specific challenges:
Function name conflicts — Internal function names like updateFrame collided with Robotics Toolbox. Solution: prefix all functions with dp (e.g., dpUpdateFrame).
Timer-based animation — MATLAB's timer callbacks require careful state management via fig.UserData.
Robustness — Users can interact with sliders mid-animation. Every callback needed try-catch protection and proper pause/reinitialize logic.
Graphics cleanup — Changing pendulum count mid-run left orphaned graphics objects. Fixed with explicit cla() and handle deletion.
Key insight: The HTML5 prototype's logic translated almost directly, but MATLAB's callback architecture demanded defensive programming throughout.
Phase 3: MATLAB App Packaging
The final .m file was packaged as an installable MATLAB App using matlab.apputil.package(). This creates a .mlappinstall file that:
• Installs with a double-click
• Appears in MATLAB's APPS toolbar
• Can be shared with colleagues who have MATLAB
Files Produced
double_pendulum_chaos.html — Standalone browser version
doublePendulumChaosExplorer.m — MATLAB source code
DoublePendulumApp.mlappinstall — Shareable MATLAB App installer
Takeaways
1. Prototype in HTML5 — Fast iteration, instant feedback, platform-independent testing
2. Port to MATLAB — Same physics, different UI paradigm; expect callback/state management work
3. Package as App — One command turns a .m file into a distributable app
The whole process — from first HTML5 sketch to installed MATLAB App — demonstrates how rapid prototyping in the browser can accelerate development of production tools in specialized environments like MATLAB.
About Discussions
Discussions is a user-focused forum for the conversations that happen outside of any particular product or project.
Get to know your peers while sharing all the tricks you've learned, ideas you've had, or even your latest vacation photos. Discussions is where MATLAB users connect!
Get to know your peers while sharing all the tricks you've learned, ideas you've had, or even your latest vacation photos. Discussions is where MATLAB users connect!
More Community Areas
MATLAB Answers
Ask & Answer questions about MATLAB & Simulink!
File Exchange
Download or contribute user-submitted code!
Cody
Solve problem groups, learn MATLAB & earn badges!
Blogs
Get the inside view on MATLAB and Simulink!
AI Chat Playground
Use AI to generate initial draft MATLAB code, and answer questions!
