> One thing we already know is that the firmware is from Apple and is encrypted, so there's no legal download.
wInd3x can decrypt it.
> and the boot ROM cannot be read from the running device
Yes it can, you just need to turn on the right clock gate.
I'm so tired of AI slop. We've been reverse engineering these SoCs slowly and carefully for years, but this is what ends up getting published and shared.
This really doesn't sound right. I've ran these from hundreds-of-watts industrial PSUs without any voltage sag and they did great. I suspect the clanker is making some shit up here, as always.
More accurate credit for the RE effort would be 'q3k took scans done by someone else for the first board and turned them into a netlist, then other people did the rest'. It's really just hosted under my GH user at this point, nothing more.
Yeah, I also feel 'just use SQLite [no matter what]' is just the pendulum swinging hard after the 'just use mongodb [no matter what]' of yesteryear.
It's so sometimes just performative. I remember when Tailscale had a similar performative approach with 'just use a JSON file on disk'. Then etcd. Then SQLite. Like sure, you can keep picking the absolite mininum technology for your needs and then change it every couple of years... Or you could just immediately go with a solid Postgres (or Postgres-like setup, eg. yugabyte) and save yourself a bunch of faffing about with weird solutions and migrating between them. But I guess that doesn't drive engagement on your blog.
There's also the more fundamental issue that Postgres - unlike something like Yugabyte - does not use a distributed consensus algorithm for writes and can lose committed writes during network partitions.
But of course, neither does Tailscale's weird DIY contraption.
Because it's not the correct term. And that's not the only thing that's wrong, see the parent comment.
With all due respect, I would recommend you run any educational resources and infographics you create past an actual subject matter expert (not an LLM!) before release. Otherwise you're just spreading shiny misinformation.
Yet you didn't say what is wrong. What is the purpose of your comment? Might not be photographed but projected through a photomask. Does it change the semantics? No. This is not a guide for developing a fab, just to be clear.
I told you - it's not the correct term. Finding the correct one should be something you can achieve in 5 minutes of research, and then you can confirm your understanding with a SME.
To be clear, I am also not an expert - I can pick out things that are wrong because I generally know a thing or two about semiconductor fabrication, but I'm not willing to sign-off on correctness of the whole thing, especially wrt. to modern processes. I personally wouldn't publish anything about this matter without getting it checked by someone else, because I know where the limits of my knowledge are.
The goal of my comment was to bring that to your attention - that you need someone to actually review your content before you continue teaching people the wrong thing. When you're educating people, you're in a position of power that comes with the responsibility of making sure what you write is as correct and clear as possible. You have a duty of care.
So you posted a comment, then refused to give details because I can do it myself "in 5 minutes of research"? Like I said previously, this is not meant to be sent to TSMC so they know what to do when opening their new fab in Arizona.
The point is that it's not a credible resource, and one or two point corrections from randos who can see some of the errors are not likely to change that.
they haven't pointed any corrections, just complained that something is wrong, which it isn't, it's just wording preference. You nailed it: just two randos on the internet making a statement, nothing that challenges the accuracy of the info.
q3k, you don't know anything about wafers/chips, you are just triggered by AI usage in all your comments.
Please give constructive feedback instead of, this uses AI and its just wrong.
I've refurbished multiple PVD systems, I'm setting up a DIY e-beam lithography rig for Cr-on-glass photomask making and I'm bringing up an 80's photolithography mask aligner / exposure system. I've even built some terrible Al/Si(P-type) Schottky junctions already. I'm not completely clueless.
From where do I know the name Dragon Sector and q3k? You aren't the ones who hacked the train DRM, are you? Or maybe active in the demo scene? Or maybe I'm just confusing you with TRSi?
sure, but s3 is far easier to buy and also far more reliable than typical distributed systems coordinator stuff like etcd/zookeeper/consul. every major cloud offers it, and there’s several decent open source options.
I haven't deep-dived into celld. Does it requires s3's strong consistency to be correct? or is any s3 api compatible object storage sufficient even if they lack strong consistency guarantee?
The fact that it's slop. People know it when they read it.
Put Claude and other LLMs away (for coding, writing, everything - block them on your firewall fully) for a few days and then write a blog post from scratch using your own words. No asking for help, comments, editing, nothing. Then compare against this. It'll jump out at you.
> Put Claude and other LLMs away for a few days and then write a blog post from scratch using your own words. No asking for help, comments, editing, nothing.
Ok, but (and this is the whole point) where _should_ we use LLMs? Because your comment could easily be said to someone writing code.
If slop === LLM involved, that's too reductive.
If LLM does all the work, that's also too reductive.
Because code is instructions to execute a program. It's not meant to be consumed as an end product itself, only its output, which is independent of what it's made from, and if written functionally correct, irrelevant for using the software.
Articles are written to be directly consumed. That's why the writing matters there.
The next question might be, why do people prefer reading human writing to AI writing. There are a few reasons.
One is that because the person writing it is an actual person, they've imagined how the reader will take the content in and have made sure that it reads well. Different people will have different writing styles and you may prefer some over others, but generally they'll make it interesting, engaging, not ramble on, respect the readers attention, get to the point quickly. AI tends to put in a bunch of filler and not really make much of a point, or say something obvious as a conclusion.
A second reason is that people want to feel connected to another person when reading something written by them. They might trust the person and build a mental image of them as they read, and let that further inform whether they agree with their opinions and recognize biases, and refine that image of the person to better understand if they can trust the opinions and perspectives in their writing. You can't do that with the output of an llm.
In summary, LLMs are good for asking questions to when researching facts, producing functional output that is either right or wrong. They are also useful but not completely trustworthy when exploring ideas or asking opinions on things. And terrible when producing text meant for humans to read (including its own output a lot of the time in chats), and any kind of visual design beyond simple templates.
I do believe there's a tiny subset of people who are more productive with AI, in the same way that Erdős was more productive with amphetamine.
But most who keep on going about their 10x productivity gain are indistinguishable from that one obnoxious guy at a party who won't shut up about his Ayahuasca retreat last spring. And they think they're Erdős.
this is a great metaphor precisely because (outside those with paradoxical stimulant response) everyone thinks they're more productive on amphetamines and there's some naive evidence to that effect. the number of things you've actually done is significantly higher. the change in value created, as measured by the decrease in distance between where you are and the the goal you're shooting for, is unpredictable at best and slowly degrades as reliance increases.
AI is like having a junior dev with an adderall addiction and an encyclopedic knowledge of coding syntax at your beck and call.
wInd3x can decrypt it.
> and the boot ROM cannot be read from the running device
Yes it can, you just need to turn on the right clock gate.
I'm so tired of AI slop. We've been reverse engineering these SoCs slowly and carefully for years, but this is what ends up getting published and shared.
reply