Unrelated: the frontpage says:
> the root of the name [Akritai] is the same word that gives us critical — which is exactly the software this effort exists to defend.
This is pure corporate slop feels good bullshit generated by an LLM. “critical” comes from “kritikós” which means “related to judgement”. “Akritai” comes from “akron/ἄκρον” which means border.
To be fair the article doesn’t sit well with me on its own, but making crappy, etymologically-untrue claim? Not on my watch.
Absolutely insane how people could have bought a bicycle that will become (partially) useless if they don't connect it with a mobile app. Even if this bike have the build quality of a spaceship I wouldn't even touch it with a stick.
Kudos to Seth for cracking the control on the bike, just so we can reclaim control of an appliance that we paid for with our own money, one that won't work because the maker can't be arsed enough to make it work without a mobile app.
Imagine after mowing your own lawn with a petrol-powered lawnmower, you got some shades from the neighbors for "Robbing the kids in the community a chance for some honest labor. If you don't know how to use the gardening shear, ask the kids to do it instead of using the automatic lawnmower. Would be a win-win for everybody".
Except with an automatic lawnmower you’re still doing the work yourself. In the video he literally said he left Claude overnight. He did no reverse engineering and no coding
Now instead of making the comparison with an automatic lawnmower, what would you say if I use your argument when you let a roomba vacuum your room unsupervised instead of asking the neighbor kid to sweep it a broom? This is why it is going to get real philosophical if we want to judge how lazy a person is based on the level of delegation he put on a tool.
The guy is not a software/embedded engineer. He's a bike engineer. He doesn't want to code more than needed to get his idea working. As for reverse engineering, he did plenty of it. He soldered the serial interface into the board, fired up a serial terminal, watch the output & note down the data sent through the wire for each event on the bike. Two things he used LLM for:
- Work through the decompiled app & find the commands for controlling the bike
- Write an app to run on the CYD.
What would be the things he would have needed to do face if he had somebody work on this project instead of Claude:
- Find somebody in his close vicinity, and have him over the workshop everyday, because this project requires access to the hardware.
- If he instead found somebody else with another Revo bike, they would have to collaborate really closely & replicate what each other have found to the other person's setup.
All that for a vanity, 1-person project?
Good grief, hopefully in v28 Lemay will also throw away the absolute crap that is Liquid glass. Aesthetically it looks horrible, it slows my iPhone 13 — which never before lagged in all its life since iOS 15 – to a maddening crawl.
Liquid glass should be taught in design school as an example of what not to do when you design UX. And also in business school as a case of how middle management can fudge up something that is working normally in the illusion of progress.
My iPhone 13 performs better since the Liquid Glass update. Animations seem to have been optimized and are much smoother. I also replaced the battery last year, so it might be worth checking if you are getting degraded performance due to a old battery.
It was still ugly, but it did kinda work because it only affected the window chrome, and it could be set to opaque with an easy-to-find setting.
On macOS the checkbox for "Reduce transparency" is hidden away in the Accessibility settings panel which is the last place I'd look for configuring the desktop theme.
Liquid Glass made my 13 mini consistently lock up and overheat. Multiple reboots every day. Ludicrously slow screen repaints. It only stopped with the latest version (26.5).
Anecdotal evidence is not proof, and as I’ve seen two people in this thread saying the performance of their iPhone 13s improved after iOS 26, I don’t see how you could concretely attribute any of your issues to Liquid Glass specifically (especially since you say it has been resolved with 26.5 despite Liquid Glass still being present in the os).
Not 13 but 13 mini, and it was fine until iOS 26. The symptoms lessened in 26.4 and now seem to be gone. So maybe the "mini" display factor was not sufficiently tested before iOS 26 was released and... it ran into snags (to put it diplomatically).
It was srsly crazy stuff. Screen freeze for a minute, for two minutes, while the battery gets hot enough to cook an egg.
A gazillion-dollar company not throughly testing their software - who'da thunk it ?
Why the hell do people keep assuming that whats-his-face was the one singular person responsible for Liquid Glass? How do they imagine Liquid Glass happened? By magic? Unbeknownst to any person in the company? Including the people who spent two hours on stage last year praising it?
Steve Lemay was a driving supporter of liquid glass, just like at least 2 other remaining designers at Apple. It's not going anywhere. Hopefully it will get toned down several times over though.
It is being toned down, at least in that they seem to be lightening up on that stupid lie that it helps you "focus on the content" because there were no toolbars. As if you're reading any text behind the buttons that are warping it and displaying another layer of text or an icon on top of it. And they ended up having to do an ugly fade at the top and bottom anyway cause you couldn't see the buttons on certain backgrounds.
I think Liquid Glass is inherently bad. Universaly accepted truths about UIs that Liquid Glass violates:
* Consistent look across instances
* Consistent look as surrounding content changes
* Contrast is meaningful
* Interactsble elements are not stacked on the z axis
* Animation conveys meaning
You can see evidence that even Apple understands this by the presence of the options to disable many aspects of Liquid Glass living in the accessibility options.
Perhaps they can add a ‘macOS Classic’ option, in the style of Windows letting you show 98-themed UI, and the many attempts at bringing back an XP-ish look to modern versions.
I’d pay… 30€ for my Mac to again look like the cat-named versions.
that's just sunk cost fallacy. The icons are also publicly available and can be used on projects conforming to the Apple branding, so it's not truly a waste.
If anything, the waste would have arguably been putting people to work on a feature that needed to get removed in the first place. If there's any post-mortem to be done, it's "why did this get released in the first place, and how do we make sure we don't spent time on something like it again?"
<critique>
OP keep saying the backend implementation doesn't matter, it can be trivially switched to something other than Github. It's not. The whole app you write revolves around Github's HTTP endpoints. There's no actual provider-agnostic Git operation in the code (for that you need to use isomorphic-git [1]). So a much more apt title for this should be: "Macaroni - a chat UI in a single HTML file that stores messages data in a Github repo"
</critique>
I'm sorry to say this, but this whole thing reeks of vibeslop, and not just the code. The docs, the readme, even the replies of OP are at least Claude-assisted, if not fully Claude-generated.
And I'm trying not to be derisive about LLM generated output anyway. LLM generated output can be sane and meaningful and to the point. This is just a questionably-cool tech demo dressed up under a sensationalistic title.
You're quoting me back at me. The quote is still correct. The backend doesn't exist. The messenger works. If that breaks your mental model of how software should be built — good. That was the point.
It's still predicting the next word. Somewhere in the gigantic dataset that the LLM was trained on, there is a phrase that says "gradient border" being in the vicinity of a CSS code that render the stuff. Therefore when you run it on an inference loop there's a good chance it output that CSS code when you tell it to render a "gradient border"
Multi-modal models that can understand visual input do exists, but no such visual reasoning process happened in the example you mentioned. Not unless you have a visual feedback loop in the coding harness.
I'm not dismissing the capability of "predicting the next word" however. The vast amount of training data enable extremely complex and useful behavior you just described.
For instance I’ve written a few custom languages to learn how to write a VM and the lexer/parser/compiler/etc. that it had never seen before and then just gave it the syntax which is different than what it had ever seen before. Simply due to the fact I made it and it had never been trained on it.
After giving it my documentation, it was able to write the language just like a language that it had been trained on. I’ve also seen this behavior at work where there are weird quirks to do things and definitely not standard and it can handle it.
Because in its training data there is information on how to map from documentation of a language to actual programs. This means that following the pattern it can map between documentation for any language to programs in that language.
But I think it will have difficulty in crossing paradigm boundaries, by simply using documentation.
That’s because it does not encode words or keywords or anything like that. It encodes their relationships. A formal language like a programming language are pretty compact. There’s not much variation between the C-like languages. Just like most Lisps (clojure, scheme, elisp, racket) are fairly similar to each other.
The exact syntax does not matter, only the grammar. If you give it the grammar, and then the keywords, it can find something that has similar grammar and then use your keywords.
I'd be very careful assuming something is not in an LLM's training set. Those data sets are truly vast. And, from experience, people tend to miss a lot of their content.
As a for instance, back in the day some academics wrote a paper that compared GPT 3.5 to a couple of inductive programming systems (including one of mine) on solving programming problems in a certain well-known esoteric language which I shall call "L". The task was to solve those programming problems one-shot. The authors asserted that the "L" problem sets were unlikely to be in 3.5's training set, but I found them without much search in a public github repo. I mean the entire dataset was right there. In this case the researchers are colleagues and friends and I know they weren't simply negligent or malicious, they just missed the fact that their "unlikely to be in the training set" data was on the web.
So I'd always assume that if an LLM can perform a task that's because it's seen examples of the task during its training.
Without forgetting that LLMs have this really shockingly powerful ability to interpolate between examples and they can improve their performance on say Task A by training on Task B, where A and B are different but similar.
e.g. they seem to get better at translating between language pairs of which they have few examples of parallel text by training on other pairs of languages for which they have more parallel text; they seem to learn something about language translation in general by training on more examples of translation. I haven't got a good reference on that handy but it's well-known (and of course over-hyped and exaggerated by tech CEOs).
So without wanting to diminish your work, I'd guess that your new language's syntax is different and novel but everything else about it is more ordinary and the similarities are such that an LLM can wing it and write you a lexer etc. After all, the whole point about parser generators and similar tools is that the task can be abstracted and separated from syntax in the first place.
In fact LLMs are very good at that sort of thing, filling in the blanks as it were. I'm old enough to remember the excitement about GPT 3.5 being able to form syntactically correct sentences with nonsensical words give to it.
For example, I just asked Chat [1]:
Hey chat. The gostak distims the doshes. What happens to the doshes?
I was building an ergo split keyboard and was also interested in this problem. After a bit of searching on the internet, I found this:
https://github.com/vlukash/corne-trackpad
The author used a BlackBerry trackpad. In his blogpost he showed that it can be mounted on top of a keycap. I believe that you can 3d print a special keycap integrate directly with the trackpad.
This is pure corporate slop feels good bullshit generated by an LLM. “critical” comes from “kritikós” which means “related to judgement”. “Akritai” comes from “akron/ἄκρον” which means border.
To be fair the article doesn’t sit well with me on its own, but making crappy, etymologically-untrue claim? Not on my watch.