MicroPythonOS graphical operating system delivers Android-like user experience
cnx-software.com208 points by mikece 4 days ago
208 points by mikece 4 days ago
Interesting.
https://github.com/MicroPythonOS/MicroPythonOS
https://fosdem.org/2026/schedule/event/9GGXNF-micropythonos-...
Indeed much better resource than this ads ridden page…
Say goodbye to most ads with Brave or Vivaldi, with adblockers enabled (EasyList, EasyPrivacy, ADB+, Fanboy's Annoyances List).
I prefer viewing the text-only version of this HN site, it loads fast and clean with no ads: https://news.ycombinator.com/
So, am I right in assuming that ESP32, being simple and slow, isn't going to have cache lines or anything, and would just need 1-2 cycles to access its RAM? In which case a pointer-chasing dynamic language like python wouldn't have all of the typical performance penalties from constant cache misses?
EDIT: upon further research, I think the above assumptions are more or less all wrong, starting with the "simple" part. To start with, they're Harvard-architecture-ish with separate memory pathways - and caches - for data and instructions, so off the bat they have more heterogeneity than your modern general purpose CPUs. Also there seems to be a very wide variety of memory mappings, buses, and caching systems within ESP32 "family". [1]
[1] https://developer.espressif.com/blog/2024/08/esp32-memory-ma...
ESP32 is still much better than this,
https://en.wikipedia.org/wiki/Amstrad_PC1512
Which was my introduction to PCs, playing Defender of the Crown at the school computer club.
While cool, it certainly won't be delivering Android-like performance, due to the lack of JIT and AOT compiler, coupled with PGO metadata, shared via the PlayStore across devices.
Micro python is the last hope for Python. Python simplicity got destroyed by a bunch of new wave of programmers who packed a lot of new useless features into it in the past 10 yrars. Now it's NOT easy and small language as it used to be...
Feature creep is an awful side effect. I would love to have language having just few add-ons per decade so I can grasp it all
I know Python since version 1.6, it was never simple, people that don't read the reference manuals are the ones that think that is a BASIC like language in complexity, when in reality it has always been more into Common Lisp / C++ kind of sense.
I still do not understand how Python won out over lua. Which is imho the superior scripting language just constrained by a smaller userbase. Lua is like embeddable in what 500 KiB. While python takes its about 200 to 300 MiB
>> "the last hope for Python"
Python is in the top 3 programming languages in the world.
This is kind of a strange take to me given that Python is quickly becoming the default for many projects that 1) are not indexing for speed/efficiency and 2) is not on the web (and sometimes this only applies for frontend). There are plenty of cases where that statement is incorrect but I think you get my point.
I think I read a title on HN that was literally titled “Why Python Won” in late 2025.
Which useless features?
Amazing. We have actually gone full circle reactionary on the typing debate where duck typing is considered the "traditional" way by some.
Walrus operator, match. They're just syntax sugar, but expand the vocabulary for little benefit.
All of them. Starting with syntax changes or type hints.... (Python should be always and only be duck typed forever as designed by God itself (it's creator).
> Python should be always and only be duck typed forever as designed by God itself (it's creator).
Isn't Guido the one who came up with type hints spec and made the reference implementation (Mypy)?
async is the big one. it was half-baked
Yeah, async Python is one of the top 2 things that I just get pissed off when I think about it too much. Along with Lin-Manuel Miranda not being involved with Moana 2.
Wow, these preassembled ESP32 plus touchscreen boards are extremely cheap, and there are tons of them in all kinds of different form factors on Amazon. I didn't realize this kind of thing was so plentiful, this seems like a great way to bootstrap many kinds of electronics/IoT projects
Yeah ESP32 is an awesome rabbit hole. An esp32-c6, cheap yellow display, and a 3d printer and you can build some really interesting things.
Any commercial products using ESP?
Just look for ESP32 CYD - CYD stands for cheap yellow display. There are a lot of variants.
https://github.com/witnessmenow/ESP32-Cheap-Yellow-Display?t... . I bought mine for about $12 and it's been quite fun tinkering with it.
https://templates.blakadder.com/esp32.html
Here’s a list of just a few. They’re insanely popular not only because they’re just good to use, but also because they’re one of the cheaper FCC approved modules you can buy, which takes a lot of the pain out of bringing a product to market.
A lot of Shelly devices use ESP chips: https://www.shelly.com/ - And they are hackable!
I think there are plenty using espressif chips. One of my robot vacuums (possibly the Neato?) certainly appeared to be.
Yes, many. As a random example, see: https://www.servethehome.com/ubiquiti-flex-mini-2-5g-review-...
The last image on the page shows various chips in the switch, the top left is an ESP32.
Python code and resource-constraint device don't seem like a good combo in terms of performance. Is it just for prototyping?
Only in the same sense as interpreted BASIC in 8 bit home computers, with their 64 KB RAM, single digit MHz, tape loading.
Good for prototyping and introduction to programming, then you had to switch to Assembly for anything meaningful.
Here, you might switch to Assembly like in those days, or maybe C, C++, Rust, Zig, whatever.
Really cool. How can this be fast? I'd have thought that Python (even MicroPython) would just be unsuitable for certain things like the graphics and animations, at the very least.
Love me some MicroPython. Building a product line of small farm security devices that use uPy and MQTT.
Would you mind elaborating on what you mean by small farm security device?
First product is called SecureCoop, that either opens the coop door or monitors another coop opener, then sends phone notifications if there are issues. Door didn't open on time, door is stuck, etc.
Next product will be battery security cameras trained on predator models (raccoon, fox, snake). Then the next one will be an electric fence monitor. All solar powered.
Basically going for back yard security for small hobby farms.
Biggest hurdle has been that I only knew 5% of what I needed to know to do this. Haha. Many mistakes. But now am close to visiting the FCC lab for SDoC verification, and then I can legally sell.
Thanks for asking!
Looks better than any Python GUI framework I’ve seen..
I reckon you've never seen flet.
That looks interesting. I had not heard of flet.
How do you like it? How easy is it to work withe the layout controls?
It's a mixed bag, as it's still not stable (esp as very recently declarative support was added in what was likely a mostly-rewrite). But when it works, it works great (I've only tried on Linux and Android).
Would it be possible to slap this on top of FreeRTOS?
Interesting. Would want to see this going on actual Android. Especially since I have a few Python GUI projects going which I intend to use on Android (but currently using flet).
It’s LVGL based, if the GUI and widgets are what you wanted you could use that on Android, although if you have access to native Android this actually doesn’t seem like the best approach to me.
The primary goal I've been seeking for a while now (and which so far only flet has reasonably fulfilled) is the ability to dynamically create GUIs. Like I created a REPL app which works similar to the regular Python REPL, but also if the code entered at the prompt returns a control/component, it shows that control with full functionality. No need to compile anything, esp on another device. And it's a standalone app, not a frontend to a server or even a web view.
Does it run on the CYD? https://github.com/witnessmenow/ESP32-Cheap-Yellow-Display
No, it says my ESP32 is not supported. I'd love to get it to run there, though.
I would love to have this, but Lua not Python.
I wonder how hard to port this to eLua would be? The bigger problem is probably packing.
Pip can work fine with MicroPython. LuaRocks can be a pain, even on desktop.
Mathematicians don't build GUIs, and nobody else can stand starting their arrays with 1.
Lua also let's you start arrays at 3.
As does Perl with `$[`[0][1]
[0] "This variable stores the index of the first element in an array, and of the first character in a substring."
[1] With the caveat: 'As of Perl v5.30.0, or under "use v5.16", or "no feature "array_base"", $[ no longer has any effect"'
You can start your arrays in Lua at 0. Conventionally you don't, but you can.
Does it support the threading module?
Will MicroPythonOS also work with CircuitPython?
CircuitPython docs > Differences from MicroPython: https://docs.circuitpython.org/en/latest/README.html#differe...
Also, there's pipkin: https://github.com/aivarannamaa/pipkin#pipkin :
> Tool for managing distribution packages for MicroPython and CircuitPython on target devices or in a local directory.
> Supports mip- and upip-compatible packages, and regular pip-compatible packages
Hopefully - for 3 types of packages - pipkin supports GPG signatures, PyPI's TUF, and/or sigstore attestations like pip?
Just checked; pip doesn't support checking PEP740 attestations yet either?
pipkin: https://github.com/aivarannamaa/pipkin
trailofbits/pip-plugin-pep740: https://github.com/trailofbits/pip-plugin-pep740
I'd use it. I'd be curious to see how close to daily driving it is for stuff like calls, SMS, and email. Something not driven by a giant data mining company would be splendid.
[dead]
> Android-like user experience
so crap. No inovation those days.
I really wish people would stop trying to innovate with user interfaces. In a comment below you criticize this UI because it doesn't have delimited interface elements. I agree that non-delimited user interface is really bad, but I attribute that mostly to Microsoft's flat design innovation, which I didn't like at the time, and I still wish I hadn't had so much influence.
As for invisible scroll bars, again we agree. But I think that was Apple. I'm sure somebody will correct me if it wasn't.
That "Android-like" is based on LVGL which is a brilliant GUI framework for ESP32 (not invented for this project) when you consider the low capacities of the hardware and how efficiently it pulls the animations.
If Android had such GUI, it would be a heck lot faster and drink less energy.
It's FOSS, so you can use it primarily for output with real switches and knobs for input. But then just using plain LVGL would probably be more practical.
I mean I kind of get your frustration, but I don't think innovating the user interface is not really the goal of this project, the opposite actually, it's moreso trying to provide a well-known user interface to devices where that was previously hard, so the goal is to be similar.
I would like to see some fresh ideas in UI though, everything is the same nowadays... :(