Hacker News: jmagland

New comment by jmagland in "SugarTrack – an offline Android logbook for blood sugar (no account, no cloud)"

jmagland — Sun, 16 Aug 2026 00:15:50 +0000

There should be more simple, no login, apps like this

Show HN: Matlab syntax in Jupyter, client-side

jmagland — Thu, 23 Jul 2026 14:01:17 +0000

jupyterlite-numbl-kernel is a JupyterLite kernel for numbl (https://numbl.org). It demonstrates running MATLAB syntax in Jupyter, client-side in the browser (no MATLAB or Octave needed). The live demo is a static site hosted on GitHub Pages... nothing to install:

https://concept-collection.github.io/jupyterlite-numbl-kerne...

Numbl is a lightweight (TypeScript + wasm + optional native bridge to OpenBLAS) numerical computing environment that tries to be as compatible as possible with MATLAB syntax. It has 500+ built-in functions and runs both in the browser and on the command line. I am developing it with Dan Fortunato at the Flatiron Institute.

Why do this? There are many codebases that are written exclusively in MATLAB, require a MATLAB license, and cannot be run client-side in the browser. One such example is chebfun (https://www.chebfun.org/), whose heavy use of MATLAB's object-oriented features means it does not run on Octave. Right now, numbl is mature enough to run most of Chebfun. The capabilities will expand over time as we implement more builtins.

I make a lot of static web applications (many of them proofs of concept) and I love trying to get code to run client-side in the browser. I think it's great for teaching, sharing ideas and methods, and communicating science with interactive views.

Caveat: numbl of course is not MATLAB. It covers a large, tested subset of the language though.

Kernel source: https://github.com/concept-collection/jupyterlite-numbl-kern...

Numbl source: https://github.com/flatironinstitute/numbl

Comments URL: https://news.ycombinator.com/item?id=49021763

Points: 3

# Comments: 0

Show HN: mip – the missing package manager for MATLAB + MEX

jmagland — Tue, 07 Jul 2026 16:36:25 +0000

Article URL: https://mip.sh

Comments URL: https://news.ycombinator.com/item?id=48820197

Points: 5

# Comments: 0

Show HN: Figpack science viz – Python script → Shareable HTML Bundle in cloud

jmagland — Fri, 24 Apr 2026 10:39:27 +0000

Sharing interactive scientific visualizations is hard. I don’t want my colleagues to have to install specialized software. I just want to email them a link which they can click on and see the full interactive visualization and explore my (potentially large) dataset and results. I want figures to be long-term archivable (Zenodo, etc). And local first (no upload required to view).

I created Figpack to solve these problems by creating self-contained HTML bundles for rich interactive scientific viz.

How it works. You import figpack in your Python project, pass numpy data (large time series, images, domain-specific data) to a figpack view object (many existing types, extensible via custom javascript/react), and then the show() command creates a completely self-contained HTML bundle with rich interactive visualization in a temporary directory and spins up a local web server to view it in the browser. The data and rendering code are all in one stand-alone directory. Easy!

To share with a colleague, just set upload=True and it uploads to the cloud. You get a url to send to your friend and they can see the exact same thing without any login.

I know… how is this different from Plotly, Bokeh, Observable, etc? Happy to discuss!

Please try it out and let me know what you think!

Open source (Apache 2.0).

Comments URL: https://news.ycombinator.com/item?id=47888312

Points: 1

# Comments: 0

Show HN: Too many browser tabs! I solved it with a simple electron app

jmagland — Tue, 21 Apr 2026 15:46:45 +0000

There are a few sites I visit very frequently and I find myself always opening a new tab. I'm used to an IDE where if you click on a file on the left it takes you to a new editor, and if the editor already exists it just takes you to that. So I thought I'd make something similar for a web browser. I admit I used AI. One hour well spent because now I don't have the too-many-tabs problem anymore.

Maybe this solution already exists? I couldn't find it.

Enjoy! And if you decide to tweak it, please submit your work back in a PR! Or of course you're welcome to fork it. Please share your improvements!

Comments URL: https://news.ycombinator.com/item?id=47850451

Points: 4

# Comments: 2

New comment by jmagland in "Show HN: Git Push No-Mistakes"

jmagland — Tue, 21 Apr 2026 15:03:39 +0000

I like the idea of having a proxy git server between you and the remote, I've never seen that before. Are there other use cases you imagine for such a setup besides intercepting and reviewing pushes?

New comment by jmagland in "Taking a Look at Compression Algorithms – Moncef Abboud"

jmagland — Fri, 17 Apr 2026 14:34:25 +0000

I've found that Asymmetric Numeral Systems (you mentioned it briefly) is the optimal practical method for pure entropy encoding. I just posted this https://news.ycombinator.com/item?id=47806122

New comment by jmagland in "Show HN: simple_ans – Asymmetric Numeral Systems Compression in Python/C++"

jmagland — Fri, 17 Apr 2026 14:11:48 +0000

Oh, you may also be interested in this fun

ANS Visualizer

https://visposts.github.io/ans-visualizer/

Show HN: simple_ans – Asymmetric Numeral Systems Compression in Python/C++

jmagland — Fri, 17 Apr 2026 14:10:04 +0000

The Asymmetric Numeral Systems (ANS) algorithm (Duda et al, 2015) is perhaps the most practical way of getting near optimal compression ratios for independent and identically distributed random sequences of symbols from a known discrete probability distribution. Simplest example: a random sequence of 0’s and 1’s with probability p of getting a 1. Shannon’s entropy formula gives us the expected compression ratio for such a sequence, but realizing that compression ratio efficiently in a computer program is not such an easy task. ANS does the trick and is incorporated into several general purpose compression algorithms, but I wasn’t able to track down a simple, self-contained implementation that was reasonably performant.

So I made simple_ans, a straightforward Python package that uses a small, yet efficient, kernel of C++ code (few hundred lines of code).

If you want it even simpler, there’s also a pure Python implementation in the repo (much slower though).

I hope you find it interesting and/or useful!

Comments URL: https://news.ycombinator.com/item?id=47806122

Points: 3

# Comments: 1

New comment by jmagland in "Show HN: I built a music theory course with games and spaced repetition"

jmagland — Thu, 16 Apr 2026 22:43:47 +0000

Yes the FBG-101 works after a few seconds. However in the original link it still doesn't work... I tried chrome and firefox. But here's a clue, when I pluck the guitar string several times, I sometimes here a little short scratch/static sound.

New comment by jmagland in "Show HN: Online Sound Decibel Meter"

jmagland — Thu, 16 Apr 2026 20:40:42 +0000

Seems to work. What do you imagine is the main use case?

New comment by jmagland in "Show HN: Modulo – A daily spatial reasoning puzzle"

jmagland — Thu, 16 Apr 2026 19:55:06 +0000

Oh tricky. Thanks!

New comment by jmagland in "Show HN: I built a music theory course with games and spaced repetition"

jmagland — Thu, 16 Apr 2026 18:28:08 +0000

The air particle visualization is really cool! However, I don't hear anything. I'm on chrome browser in linux. It wasn't clear how to get that to work. Thx

New comment by jmagland in "Show HN: Arrow. Point your phone, walk where it says, find out where"

jmagland — Thu, 16 Apr 2026 18:20:08 +0000

Fun idea. Before I try it, how far should I expect to go before reaching my destination?

New comment by jmagland in "Show HN: Modulo – A daily spatial reasoning puzzle"

jmagland — Thu, 16 Apr 2026 18:12:44 +0000

Nice! Is this meant to promote brain health, or to just be relaxing?

Q. Is it possible to resize the shapes? That wasn't clear to me from the interface, and I had a hard time completing one of the first puzzles without being able to resize.