Programming Leftovers
-
AI Art Just Opened The Threat to Human Work We Were Expecting from AGI
No, I’m talking about major attacks on knowledge work within 5 years, with something like 50% to 80% of knowledge work is doable by AI within 8-15 years. Whether it will be done by AI that’s another story, but the capabilities will be there.
-
[Old] Embedded ML for All Developers
Now the age of tinyML has come — we can already run meaningful ML inference on Cortex-M equivalent hardware. Rapid improvements in modern 32-bit MCU compute power efficiency and math capabilities (FPU, vector extensions), together with advancements in neural operators, architecture and quantization along with better open source tooling like TensorFlow Lite Micro are making this possible. For example, we recently built a complete DSP, Anomaly Detection and NN classifier for complex events on real-time 3-axis accelerometer data in software on a standard Cortex-M4 in just 6.6 kB of RAM and 20 kB of Flash. We are experiencing the start of what I call the “3rd wave of embedded compute.”
-
In defense of linked lists
So, get ready to read a sentimental post about a data structure, and don't tell I didn't warn you.
Linked lists are educational. When your teacher, or the page of a book, or anything that exposes you for the first time to linked lists shows you this little circle with an arrow pointing to another circle, something immense happens in your mind. Similar to what happens when you understand recursion for the first time. You get what data structures made of links truly are: the triviality of a single node that becomes a lot more powerful and complex once it references another one. Linked lists show the new programmer fundamental things about space and time in computation: how it is possible to add elements in a constant time, and how order is fundamentally costly, because if you want to insert an element “in place” you have to go from one node to the other. You immediately start thinking of ways to speed up the process (preparing you for the next things), and at the same time you understand, deeply, what O(1) and O(N) really mean.
-
[Old] We need to talk about GitHub
Have a seat and make yourself comfortable, we're gonna talk about GitHub. Yes, GitHub. And I'm not referring to the ICE Contract, the performative activism, the fleeing of their women senior engineers, or the other litany of accusations against the company. I want to talk about the idea of GitHub.
And for that matter - I want to talk about the idea of SourceForge, Google Code, BitBucket, GitLab, etc. The entire concept of a single, centralized project & repository hosting service is fundamentally at odds with the health and longevity of the open source community.
-
The essence of Reed-Solomon coding
Reed-Solmon coding gives us much more flexibility, allowing us to store our data over n = k + t drives, so that any t drives can fail while still not losing data.2
-
Interview: Dr Luis Falcon, GNU Health, free software in healthcare
Dr Luis Falcon is a medical doctor and computer scientist. He combines these skills to work on GNU Health to bring free software / open source to the healthcare sector. He spoke with us about progress, ongoing problems and an upcoming event.
Ciarán O'Riordan: You're working on free software / open source for the health sector. Is the health sector lagging behind?
-
[Old] Using Python to vectorize artwork for PCBs
In this article, I want to talk about just one approach for converting artwork for PCBs- Image tracing, sometimes called raster-to-vector conversion. This process takes a raster image (such a png) and converts it into a list of polygons. This article will show how I used Python to glue together a few clever libraries to build gingerbread.trace, a tool for converting images into KiCAD footprints.