Hey Alex! Just wanted to say I came across Slicer three or so months ago, researching automatic provisioning for RPis (and other devices in the homelab). It piqued my interest, as I was also researching solutions for partitioning and distributing workloads. I hope to give Slicer a go within the next quarter, however I'm still getting to grips with the fundamentals of distributed computing. Either way, it's interesting work.
It is possible to exist in both camps. The quality of the process affects the quality of the product, and the quality of your thought affects the quality of the process. It's a cycle of continual learning, and from that perspective, thought, process and product are indivisible.
Treating code as a means to an end doesn't guarantee success for your product anymore than treating code as a craft.
It's not possible for two-camp believers to conceive of two-camp dwellers. That would be tantamount to a third, potentially superior camp.
The two-camp construct is a tool to establish the believer as a member of the supreme one camp group; apart from the lesser campers. Their entire identity and self worth is built around one-camp membership.
Quality is 2nd best to what? And people haven't switched to what? Android? The situation is no better on Google OS.
Apple's App Store ad initiatives have always been woeful, and doubt it makes enough revenue to warrant a separate line item on their public accounting reports. Some executive has seen yet another overfunded company potentially making bank with an ad-based business model (OpenAI, et al.), and has thought they could extract Google-level ad revenue due to the App Store's exclusivity. It could also be a response to potentially competing App Stores given their rocky relationship with the EU.
It will have little effect, on revenue or user experience. The greater tragedy is the organisational decay that led to this being greenlit in the first place.
> And people haven't switched to what? Android? The situation is no better on Google OS
Agree. Even GrapheneOS is hell to use. I tried both PixelOS and GrapheneOS on a Pixel 9 and ended up returning it. If I was not homeless I would switch to a flip phone and just use a Linux desktop.
This is not my experience. GrapheneOS is great and has absolutely zero bloatware/malware. The base system is just a couple of basic apps like the phone, messages app, and a Web browser. That's it. All the rest is up to the user to set up. You can be completely Google-free if you don't install sandboxed Google Play Services and other GApps.
Without GApps, the setup is extremely private and ideal to use with self-hosted solutions like Nextcloud and Home Assistant as substitutes for the typical commercial malware found on most "smart" phones.
The company cares about neither. People inside the company will care about a great many things. The people who care about users either don't have the power to act or no longer care enough to do so.
If the company was trying to extract as much profit as possible, it would be doing so at every level; it would be a company-wide strategy. This just looks disjointed. It speaks to Apple's loss of social cohesion, the signals of which have been apparent for sometime.
This isn't an "I told you so" moment, as this initiative is meaningless without context, and it's a poor attitude to take.
I don't want to take away from the work you've done, as you're clearly knowledgeable, but as someone else observed, heavy use of AI assistance can be observed in all your public projects. It's worth explicitly addressing, especially considering the foundational nature of your project: it's not easily replaced if it turns out to have to have subtle bugs.
Though I rarely use it myself, I'd like to know, simply because I'm curious as to how other engineers have incorporated such assistance it into their process.
I was interested in provisioning one of these a few months back through Scaleway, but couldn't navigate their sign-up process without it dumping me back to the start everytime. Nor did I receive a reply when I e-mailed their support e-mail.
I don't know if that's changed (they had odd pricing too, like Startup vs. Business, of which the difference wasn't clear), but aware. I hope someone has more success than I did.
Grim environment in the rodeo. The /home/directory, which the Rasberry-Pi doesn't align or distill as a symbolic markup language in .async or transport layer protocol and refers to /dev/ spinlock rolling release kernel version recursion or discursiveness.
Scaleway's site and support is horrible, as you've found. But their pricing and offerings are solid and their network is... OK. For the price, they're better than Hetzner... if you can get signed up!
Totally different experience here! For a project I wanted to try “euro cloud”. Something comparable to digital ocean. No need for hyperscaler functionality.
It has been great. Good terraform provider and reliable service. I like their console, although the design feel very vaporwave to me.
Of course stuff can be better, but it is rapidly improving. The way they implemented grafana + user management was shit. But that’s fixed. Grafana still feels bolted on however. And login is a bit weird with their dedibox or whatever button next to the cloud offering. But no where near as confusing as aws is!
Also bumped on a bug in their terraform provider, found a related bug report, contributed some info and it was fixed within two weeks.
Quite happy so far. Running serverless sql, serverless containers. Secrets management and some iam config. No big stuff but quite sure it is capable to run a decently sized saas.
I won't comment on the reliability of their services, as I've not experienced it. I was signing up specifically to provision a M1 mac mini, and couldn't navigate it. It was unfortunate, but worth a comment, in case others experience the same issue (or someone else could point me in the right direction).
I don't think wpa_supplicant.conf has been used for sometime, as they moved to cloud-init for bootstrapping. It requires the network-config file instead, the format of which is documented on the cloud-init documentation page.
I happen to have been experimenting with this for the past few weeks, and the most persistent issue was getting wi-fi to work correctly. It's quite a common issue, with any number of hacks. I offer my own network-config below, though I've only tested it with provisioning Ubuntu Server on the RPis so far (I have two 3B+s).
1. The renderer, as the default is NetworkManager, which doesn't work correctly with RPis (at least on Ubuntu Server). It may work with RPiOS, but I haven't tested it yet.
2. The regulatory domain, the lack of which is what disables wi-fi in the first place. I forget how much testing I did with the format, but I believe it must be uppercase (I don't remember about quoting the string, however).
3. Disabling IPv6 may be relevant, though unlikely. It was just in a working configurations I found; I just haven't had time to confirm it. The relevant line in my user-data file is as follows:
The rest of the configuration is standard, though I purposefully made the wi-fi non-optional so I could confirm that wi-fi worked (my only Internet at the moment is through my iPhone hotspot, which was another source of issues, but that's a whole other story).
NB. According to someone else, the imager has the respective command line options for user-data and network-config, which I didn't know.
What were some of the technical challenges you experienced while reverse-engineering the wallpaper system? I've been reverse-engineering (for lack of a better term) some of macOS' and Xcode's poorly-documented functionality prototyping a personal developer tool. My investigation isn't sophisticated by any means; it's just been trial-and-error, but I haven't found much online in the way of resources for people going down this route.
Reverse engineering is hard! I use Hopper (https://www.hopperapp.com) to disassemble related binaries and frameworks. It's a great way to explore whats actually happening within macOS or Apple apps.
You can also export assembly files and throw various agents (Gemini, Claude etc) at them to learn more. It's surprisingly effective!
I haven’t tried those, but Hopper has been working pretty well for me. Although I mostly just sift through the assembly and pseudocode, and don’t use many advanced features.
This article is yet another reminder I need to learn Haskell (I've been meaning to for a decade), although the code from this article is approachable considering the topic. However, I've just started using Rust for professional projects, so the code you've posted is a bit easier to read, if more verbose, though the concepts are still unfamiliar to me.
I'm assuming this isn't your first go at writing a compiler?
Glad someone found it useful! It's at least represents a more fleshed out working example, and it's in a little module so it's pretty self-contained and easy to read through.
> I'm assuming this isn't your first go at writing a compiler?
Not quite, the first real language I worked on was called Eve: https://witheve.com
Ignoring its negative connotation, it's more likely to be a highly advanced "stochastic parrot".
> "You don't do that without some kind of working internal model of mathematics."
This is speculation at best. Models are black boxes, even to those who make them. We can't discern a "meaningful internal representation" in a model, anymore than a human brain.
> "There is just no way you can get to the right answer by spouting out plausible-sounding sentence completions without understanding what they mean."
You've just anthropomorphised a stochastic machine, and this behaviour is far more concerning, because it implies we're special, and we're not. We're just highly advanced "stochastic parrots" with a game loop.
> This is speculation at best. Models are black boxes, even to those who make them. We can't discern a "meaningful internal representation" in a model, anymore than a human brain.
They are not pure black boxes. They are too complex to decipher, but it doesn't mean we can't look at activations and get some very high level idea of what is going on.
For world models specifically, the paper that first demonstrated that LLM has some kind of a world model corresponding to the task it is trained on came out in 2023: https://www.neelnanda.io/mechanistic-interpretability/othell.... Now you might argue that this doesn't prove anything about generic LLMs, and that is true. But I would argue that, given this result, and given what LLMs are capable of doing, assuming that they have some kind of world model (even if it's drastically simplified and even outright wrong around the edges) should be the default at this point, and people arguing that they definitely don't have anything like that should present concrete evidence ot that effect.
> We're just highly advanced "stochastic parrots" with a game loop.
If that is your assertion, then what's the point of even talking about "stochastic parrots" at all? By this definition, _everything_ is that, so it ceases to be a meaningful distinction.
