Synctrain is an open source (MPL2.0) iOS Syncthing client (which I made) with full native mobile-first UI and tight iOS integration (shortcuts, background processing, etc).
You are most likely in France: your government does not allow publishing an app containing cryptography (in this case, Golang's crypto implementations and a package used by Syncthing - only using iOS libs should be fine) without authorization (which can only be obtained through French forms, at which point I'd want a French lawyer to be involved, so no).
You could of course build the app yourself from source.
Similarly, Trixie contains an updated version of Dovecot that (even though the version number seems to indicate otherwise) has a new configuration format that is not backwards compatible. This is clearly stated in the release notes but may be surprising nevertheless.
So, basically any local productivity tool, saving files in a synced folder.
While this works, Syncthing does not really provide anything for fine-grained collaboration or sharing (you only share full folders). Encrypted peers do allow storing files on a machine that you don’t have to trust.
Any tips for an app to use on iOS to capture the necessary .ply data?
Scaniverse is a great app by Niantic that can do this on-device, but it isn't very customizable and can't export its raw scanning data (exported .plys do not have the data this editor requires).
In my neighborhood (Netherlands) it appears the fiber network is physically point-to-point (subscriber to ODF), but is operated as XGS-GPON (so all subscribers see the same light signal so to say, but each over their own ptp fiber from the ODF). So point-multipoint only at the active layer.
I was told that this is because the company who is rolling out the fiber wants to make the network as attractive as possible to ISP’s who want to offer services over it (and wants them to compete) which may be more difficult in an actual physical point-multipoint network (which requires PON). The ISP currently likes PON more than AON (basically Ethernet over fiber to a switch) because the equipment is cheaper. In theory I should be able to switch to an ISP who offers AON or its own PON (they’d only have to physically patch my fiber in a different port at the ODF).
Author here, I guess I could have chosen a better title. This article is about diving into complexity that seems insurmountable at first, with an engineer's mindset.
Author here - this figure is insane but also a bit of an exaggeration. I calculated it by running 'cloc' on the source code directory, not excluding node_modules. Many modules in there are actually unused or only used at build time. The built output (from Nuxt) that is actually run is much more compact (though partly minified so line counts are not that interesting). Also, Nuxt specifically is a bit overkill here, but greatly improved my productivity developing this website (and was chosen because it aligns with our other web applications, so less 'innovation tokens' [1]).
Not a single mention of the word ‘Apple’ in the original post, instead ‘the manufacturer’ and ‘the big corporation’.. curious if that is deliberate and if so what the reasoning is (legal?)
As they are accepting a JSON schema for the function calls, it is likely they are using token biasing based on the schema (using some kind of state machine that follows along with the tokens and only allows the next token to be a valid one given the grammar/schema). I have successfully implemented this for JSON Schema (limited subset) on llama.cpp. See also e.g. this implementation: https://github.com/1rgs/jsonformer
Also WONNX can both be used from native apps (through wgpu it uses Vulkan, DX or Metal) a well as on the web (using WebGPU, WONNX compiled to WebAssembly).
Actually WONNX also transforms to an IR first (early versions did not and simply translated the graph 1:1 to GPU shader invocations in topographically sorted order of the graph). In WONNX the IR nodes are (initially) simply (copy-on-write references to) the ONNX nodes. This IR is then optimized in various ways, including the fusion of ONNX ops (e.g. Conv+ReLU->ConvReLU). The newly inserted node still embeds an ONNX node structure to describe it but uses an internal operator.
This makes running larger machine learning models in the browser feasible - see e.g. https://github.com/webonnx/wonnx (I believe Microsoft's ONNXRuntime.js will also soon gain a WebGPU back-end).
You can indeed perform inference using WebGPU (see e.g. [1] for GPU-accelerated inference of ONNX models on WebGPU; I am one of the authors).
The point made above is that WebGPU can only be used for GPU's and not really for other types of 'neural accelerators' (like e.g. the ANE on Apple devices).