-1 it's AI coming for our laziness - 1000s of CVEs the past month from well regarded software of all kinds, even Linux issuing waves of CVEs and PostgreSQL 19 delayed for the first time in years.
-1 it's perhaps offensive but I agree with OP: 3rd order is *very* difficult to reason about, and even really smart, really experienced people have colossal failures at that level of reasoning-depth. It's not just probabilistic (e.g. chess, go), but in addition the way a decision-tree-branch is followed often deeply affects downstream decisions...
See, this is more along what I expect from a place that titles itself "Hacker News".
The reason I mentioned 2nd order and 3rd order cognitive abstraction, is because to fix the web we need to make some tough decisions.
When technical enthusiasts such as people who call themselves "engineer" but aren't real engineers hear these things the response is "nobody is going to do $thing lol". Yet the more big brains are indeed talking about doing thing or lamenting the mal-growth we have collectively experienced, including the recognized inventor of the WWW.
The below is just off the top of my head, "in passing". Some people have done some work on parts of this picture already. Maybe I could post something more "formal" somewhere later on.
Before jumping in with "omg but then how would X/Y/Z work lol", use whatever 2nd order or 3rd order abstraction thinking you might have, and consider that those things were already considered by people working on these problems (or at least attempting to) and yes there are full answers or at least a couple of half-answers for those concerns.
Among those is content-addressing. This eliminates, by avoiding outright through design, most of the Cloudflare/Akamai, "verifying your connection", and captcha bullshit. This also means user metrics are opt-in, which maybe they could be by default unless a user turns it off, but then they would be easy to spoof (which they already are).
The major thing is no more untrusted Javascript and no more untrusted WASM unless loaded in a pop-up window as "external program". See the Free Software Foundation's stance on non-free Javascript. This stance isn't so "out there" to legitimate hackers.
Due to decisions made early on with processors and the feedback loop optimizing for certain styles of programming, our CPUs do speculative execution. Some developments have helped, but the only safe option for CPUs people actually use, which are not the second or third latest generation (often for chipset security concerns), require disabling Javascript JIT and WASM or Javascript all together.
Almost all of the widgets and doodads we need for the web have been figured out by now. We need a trusted collection of libraries for these things. This should include all of the extra fonts we want.
Any GPU/canvas stuff needs to be truly deterministic so as to be useless for fingerprinting.
No more User Agent string.
No more sending screen dimensions. Yes, how to contend with this has been thought of already.
Not every website should be "mobile friendly". Yes, you do need a fucking laptop/desktop to get by in the world. "I can do everything on my phone" is NGMI.
The whole CA thing needs a complete overhaul.
Cookies need to be handled in a way that users can hover over a browser tab and see what cookies are visible to the remote end of that browser tab, and cross-domain cookies must be explicitly allowed by a user.
There are very tough decisions and things that do require serious talent to innovate correctly in a way intuitive to users, that need to happen to eliminate
fingerprinting.
It's not those incapable of 3rd order cognitive abstraction who are going to dig us out of this mess. It's people who actually give a shit and think about these things a lot who will quietly implement the fixes and will never be renowned or heard of by the crab bucket mentality monkeys unless someone in authority tells them too.
I think I agree with you, but the effort to "judge it like any low-effort crap" is now beyond anything I can muster. I think when people say "ban" the often mean filter with such an aggressive heuristic that I never see it, which IS one way to judge...
omg where do I start... that horrible video? which highlights the limits of AI not the strengths!!! and gets almost everything wrong, from sunflowers to DNA.
??? everyone I know who's gone fold hasn't gone back.
yes, battery & camera are a generation back and water/dustproof is even worse.
but productivity and fun are off the charts: slab phones feel pathetic once you've got a fold. Communication apps show you the list of conversations as well as the current one - or the current one and the details/multimedia. Google Maps shows the map and the listing details/reviews. Games and multimedia are delightful. Photos are full sized, and you can see what you're shooting in detail - if you shoot with a digital camera, remote shooting is amazing because there's enough space to show detailed focus.
Not to start an endless device debate but a good chunk of these use cases make more sense to do on an iPad or to just have a separate device in general.
I know, I know, that’s not as portable, that’s not a device you have with you all the time like a phone.
But we’ve seen pretty mixed results on the “combine devices into one” playbook. Sometimes it’s a huge hit, sometimes not so much.
The annual AppleCare bill on an iPhone Duo is almost as expensive as an unlimited T-Mobile data plan for an iPad, $240/year vs. $200/year.
You are correct in saying that yes this is one device that roughly is equivalent to an iPad mini, but in terms of cost, durability, screen quality, etc, you have to really find the iPad mini to be unworkably difficult to carry around to make the cost and fragility of a foldable make sense.
There are benefits to having two devices in addition to the downsides. Two people can use them. Two batteries versus one, double the run time.
Why is an iPad mini a nonstarter for a commute? It could fit in a purse or small bag at most. You could get a cellular plan for an iPad for the same cost as the annual AppleCare bill on an iPhone Duo.
Why do I need a big screen device for a restaurant or club again? Shouldn’t I be enjoying the restaurant or club? Wouldn’t the heaviest and most fragile phone in Apple’s lineup be a poor choice for a club?
Is it hard to learn an iPad compared to an iPhone? What learning or optimizing am I needing to do? Am I unable to sync information between my iPad and iPhone in some way?
I can stand up my phone or tablet with a $20 case.
What problem is the Harry Potter moving newspaper solving? Is that worth $2,000 plus $200/year for AppleCare so you don’t get caught paying for a $1,000 screen replacement?
It's a upstream targeted fix that was made available significantly faster in packages/images maintained by volunteers, it's not some untested downstream patch. Customers at this point would have likely taken their applications offline over this issue so it would have been difficult to make the situation any worse.
Given the sheer number of PHP applications out there, it's impossible to conclude with certainty that this targeted fix will have no regressive impact without significant testing. How much testing needs to be done is debatable, but I'm pretty sure the answer isn't zero.
Someone with a single gigabit internet connection can scan the entire IPv4 address space in less than a day. When you have public RCE vulnerabilities like this, your options are "apply the patch", "take the system offline until you can apply the patch" or "get pwned".
Mitigations and patches are in the same category, so now you're applying mitigations without testing. What if your new WAF detection signature has false positives etc.?
The difference is that a mitigation is customer-applied, not vendor-supplied. Like it or not, blame shifts when a vendor supplies a fix that's insufficiently tested. The CrowdStrike BSOD incident from 2024 comes to mind.
As for WAF rules, those can usually be flipped on and off with a single switch.
Being able to roll it back isn't really a difference, you can revert to the previous version of a package too.
The trouble is if the WAF change e.g. messes up a redirect or turns a success response into an error response or vice versa, which confuses some other system into giving multiple users the same session cookie and gives them access to one another's accounts or writes corrupt entries to the database or whatever it is. You can revert the rule but that doesn't revert the state changes that have already happened.
There is inherently a trade off between keeping the system online, keeping the system patched and testing the patch before applying it.
> Being able to roll it back isn't really a difference, you can revert to the previous version of a package too.
Sometimes. In a world in which customers are increasingly adopting immutable infrastructure patterns and (contrary to best practice) using Launch Templates with symbolic references to `latest` versions of AMIs, this isn't always easy.
> There is inherently a trade off between keeping the system online, keeping the system patched and testing the patch before applying it.
I don't think you're being understanding or empathetic here.
Any vendor that delivers updated packages without thorough regression testing that ends up causing an incident for their customers as a result will get a thrashing from the community. Remember the CrowdStrike BSOD incident in 2024?
Maybe Ubuntu and others are cutting corners to deliver more quickly. It's their risk to take, but it's still a risk.
Unfortunately people rarely follow best practices.
Customers frequently do risky things like using launch templates that refer to the latest AMI. AMI changes, Auto Scaling launches a new/replacement instance in the fleet without adequate testing, *boom*.
> Track record is extremely relevant when the "argument" is mostly about judgement and subjective opinions.
Ironic given this article is from CNBC, employer of Cramer, host of "Mad Money", who has been so inconsistently wrong that his picks have literally sparked an ETF. Just search using your engine of choice "Inverse Cramer ETF".
I work in finance. Past performance is not a guarantee of future returns, but is a damn good indicator, which is why everyone always wants to see your past performance anyways.
SJIM lost 15% in a period when the S&P gained 25% (betting against Cramer). LJIM over its lifetime gained 3.7%.
Of course both had too short a lifetime to draw a conclusion due to lack of interest and this was pre-AI so maintaining the positions was difficult. But it’s important to remember that Jim Cramer fails to beat the market, not that he’s an inverse oracle.
> Past performance is not a guarantee of future returns, but is a damn good indicator,
Unless it's about stock picking, when the winners in one year of doing good stock picking are no better in the next than any other random stock picker.
Wise to critique both holistically, which makes it funny when people want to ignore an individual's or group's track record (especially regarding specific mental models or types of analyses) when evaluating a specific argument
If a weather reporter on TV says it's raining every day, and it's not always raining, I can still criticize them on days when it rains without it meaning I don't think it's raining.
I think it's probably a negative for the stock price when the CEO thinks his company shouldn't be developing its main product and believes wiggy and false things about what continuing development means. Investors can put up with a lot of things, but an erratic willingness to undermine core business activities is usually not one of them.
Is your purpose isn't to dig ditches for another man to fill ?
When we discover and fix bugs and CVEs, that's what we're doing. It needs to be one, but humans don't have to do it.
reply