Having something Turing-complete is surprisingly easy, and it hides everywhere. The repository have a small document that explains how you can use printf() as a computer : it can performs additions, logical union and negation, which is enough.
It was unintentional, but Ken Thompson being Ken Thompson, can't be 100% sure.
Don't look at the end destination, look at the journey to the destination
* Learn low-level details of a basic but real-world CPU
* Practice the brain gymnastic of programming an atypical Turing-complete computer
Your created new connections in your brain, put to use some of the old established connections. Having a machine spit-out the emulator would rob you of all that. Like, you can drive from A to B, but running for A to B can do you much good.
Hot take : the whole LLM craze is fed by a delusion. LLM are good at mimicking human language, capturing some semantics on the way. With a large enough training set, the amount of semantic captured covers a large fraction of what the average human knows. This gives the illusion of intelligence, and the humans extrapolates on LLM capabilities, like actual coding. Because large amounts of code from textbooks and what not is on the training set, the illusion is convincing for people with shallow coding abilities.
And then, while the tech is not mature, running on delusion and sunken costs, it's actually used for production stuffs. Butlerian Jihad when
Self-plug here, but very related => Robustness and the Halting Problem for Multicellular Artificial Ontogeny (2011)
Cellular automata where the update rule is a perceptron coupled with a isotropic diffusion. The weights of the neural network are optimized so that the cellular automata can draw a picture, with self-healing (ie. rebuild the picture when perturbed).
Back then, auto-differentiation was not as accessible as it is now, so the weights where optimized with an Evolution Strategy. Of course, using gradient descent is likely to be way better.
I fixed my recurrent back pain with a 6 mn daily morning, ie. plank, side plank, reverse plank, 1mn 30 sec each.
Posture muscles are not very well known in the general public. Loss of strength due to aging and sedentary lifestyle makes standing, seating, etc uncomfortable.
Significantly faster compilation means less friction to iterate ideas, try things, which in the end lead to more polished results.
A nice interface is agreable, but maybe there are diminishing returns when you pay it with large compile time. I remember pondering about that when working with the Eigen math library, which is very nice but such a resource hog when you compile a project using it.
Be warned, zero documentation, because things are at larval stage and change often. Will include a couple of demos this week.
In the spirit, yes, but targeting different hardware, public, and environments.
* It runs on Linux, Mac, Windows. Bare metal on rp2040 and rp2350 is planned.
* It written in C, build with Make.
* It is meant to run on something like a Raspberry Pi, Latte Panda, etc
* A setup is a text file, no fancy UI.
* The plan for live parameter fiddling will be a web server. Web UI will be tailored to each setup, no one size fits all UI. Typically I pay someone to do the UI.
* For now, it's only video, no sound output
It will be used for several large interactive LED displays and object tracking systems. It's a way for me to factories all those projects I was contracted for.
A nodal real-time video processing tool : put together pre-made "processing boxes" to generate interactive video. It runs on pretty much anything, uses a plugin architecture.
Say, plug a camera, and it will blend two videos streams using a silhouette detected on the camera, with various effects. It's very, very early, pre-alpha stuff, but it already was used for a demo by a customer.
GitHub pestacle, be warned, it's undocumented and larval stage
It's using the Sun as a (gravity) lens, with probes at the focal point to gather the image. Because it's a very large lens, that's allow to have a massive zoom on whatever object we are interested in.
I had the same experience with computational geometry.
Very good at giving a textbook answer ("give a Python/ Numpy function that returns the Voronoi diagram of set of 2d points").
Now, I ask for the Laguerre diagram, a variation that is not mentioned in textbooks, but very useful in practice. I can spend a lot of time spoon-feeding the answer, I just have the bullshiting student answers.
I tried other problems like numerical approximation, physics simulation, same experience.
I don't get the hype. Maybe it's good at giving variations of glue code ie. Stack Overflow meet autocomplete ? As a search tool it's bad because it's so confidently incorrect, you may be fooled by bad answers.
Fixing the topological naming issue, in the mainline, what a game changer.
I am using Freecad for Actual Real Things. I learned to work around the topological naming issue, but it cost me time, and it can make parametric models quite brittle (ie. a minor change can break the model).
For real, because I am way more productive with FreeCAD. FreeCAD allows to work in term of topological features like surfaces, edges, etc which is, in practice, very cumbersome with OpenSCAD.
That flap is already a legend, kept at it even mangled by hot plasma, crazy accelerations and pressures, spitting molten steel at the camera. What a role model, the little flap that could.