But this is already how open source works today. If you have the code, you, a human, could find and 'fix' or exploit vulnerabilities as much as you want.
Now if Fable had an easy jailbreak like this that allowed you to attack remote targets that'd be a different story but I genuinely cannot see how neutering its abilities to 'fix' code you already have access to is sensible. It would destroy the value of the model. And don't forget, any actor not abiding by the same rules could develop an model for offensive use just fine, so this protects you against exactly nothing but does destroy a potential defense.
In the end this all comes down to legislation, in much the same way platforms are not responsible for copyright violations IF they abide by some rules, the same has to happen for AI providers. If you have a process for reporting 'jailbreaks' on illegal actions, and prevent users doing illegal stuff on a best effort basis, the rest of it should really just be individual responsibility. If a user wants to use an LLM to crack systems, fine, that's already illegal.
If Tesla FSD deliberately hit somebody, holding Tesla liable is fine. If you messed with FSD until you finally got it to hit a person, then you should be liable. Outlawing FSD because it could theoretically be tampered with is just an odd stance imho.
I know this is only tangentially related, but recalling messages is horrible. I hate that so many services will allow people to send me a message, give me a notification with a preview, but then the message gets edited or deleted. If you drop a letter in a physical mailbox, or slide a paper underneath the door, you cannot get it back either. This whole philosophy of 'we allow destruction of messages in a shared chat' needs to stop. The moment things are being sent, both sides are co-owner of that message. Not being able to recall messages is a good thing.
I'll settle for a brief edit (not retraction!) window after sending though, say 5 minutes tops.
Edit (I realize the irony): banks of course won't give a hoot about the receiver, the power dynamic is inherently not equal.
Asking Claude to commit and push triggers the Co-Authored-By thing typically, even if the change was made by hand. It could entirely be possible the author just asked it to generate a commit message for this change (although the style doesn't strike me as very Claude like).
I'm not anti right-to-repair by any means. I'm just saying that "we don't need the manufacturer, just vibe code your own firmware" is not a solution to this problem.
Farmers should be able to swap parts, tweak loads of things in software, etc without software locks. I should be able to do that on my car too. But the way to achieve that is not through spending a few hundreds of thousands of dollars on Opus [latest] and praying that doesn't kill me, that's the part I have an issue with in OPs comment.
I've never been fond of the argument that there should be a professional software engineer certification, but hearing people like you being presented with the potential dangers and just going 'oh yeah just go with a better test suite and you can just wing it' makes me seriously reconsider.
Vibe code administrative systems for your local golf club to your hearts desire for all I care, god forbid somebody will have to stand around a bit longer before going for their 9 holes. But safety critical equipment is not the place to fuck around with the code prediction machines that have existed for 4 years, have been writing more-or-less acceptable code for 2, and will still regularly refer to themselves as MechaHitler or just make up shit. "Yes you're absolutely correct, I was wrong" doesn't help you one bit if you have just been chewed up by heavy machinery, and the fact that people like you exist who go 'oh just a few more more unit tests surely will fix it' is a terrifying thought.
Aside from the legal question whether the manufacturer allows you to do so, I’m pretty excited about somebody vibe coding firmwire to a 35 ton machine with a bunch of big attachments at the back and plenty of ways to mangle the bodies of careless operators without the rpm so much as audibly rising from strain. Should give us plenty of videos to traumatize the next generation of children with a little bit too much internet access at an early age. I feel nostalgic for those days.
(This is sarcasm, pretty please don’t vibe code car firmware, let alone anything more dangerous than that)
It’s especially frustrating that dependency hell seems to be embedded in the Python culture. The amount of “oh no this lib will only work with Python 3.10+ and a slew of other libs at random versions we won’t bother to tell you” while some other lib that it depends on will only work on “3.8.56z but not if you look at it funny and only if you say pretty please” is maddening. Semver is apparently not standard practice either.
I am probably biased against Python, so take this opinion with a grain of salt, but it feels to me like a whole ecosystem of amateur software devs (but professional ML-engineers, data scientists etc) cobbling together something that barely works.
I’m old enough at this point that I remember the whole old guard of software engineers falling over themselves to hate on JS and Node, call the ecosystem immature, and be quick to point out how that is not “real” software. But in the past 10-15 years it appears JS and Node got their shit together, while Python is still completely and utterly stuck in managing dependencies and environments like it is 2012. And if you ask professional Pythonistas about this, you always get an answer like “oh it’s actually really easy, you must not have taken the time to really look at it, because Python is easy, it’s just pseudocode look how amazing it all is”
I really wish ML hadn’t standarized on Python. As a user of ML tools ans frameworks but not a fulltime ML engineer it just constant pain.
In the best case scenario, professionally writing software is treated like a craft. You produce something useful while applying your skills, with the tools at your disposal. You can write software as an art form, just like woodworking can be both a craft and an art form. A woodworker in an assembly line doing the same thing over and over again to me is not a craftsperson, it is the attitude towards the job that makes it art, craft, or assembly line work.
Too many software projects treat programmers as factory workers, where their primary value is measured in amount of storypoints or Jira tickets finished. Don't get me wrong, you can be a craftsperson and use an issue tracker ofcourse, but if quantity is the only thing management cares about instead of quality, the craft gets lost in the process. Quantity is easy to measure, quality is not.
At the same time treating software like an art is probably not very useful. That code is (typically) not written to be looked at, but to make the computer do something useful.
It's a shame artisinal software sounds so weird, because that precisely describes the level of caring I'd like to see applied to the software I use.
I don't like the superpower vibe either. I do not entirely look negative upon my ADHD, but then again, to what degree it is my personality and to what degree it is the disorder, I don't know. If others want to view their ADHD as a superpower, I'm all for it, but pushing this narrative too hard doesn't feel constructive.
For me personally, I tend to view myself as a human with some characteristics which in most contexts are not helpful, but make me uniquely suited for other situations. I try to avoid the word 'broken', but for a long time I thought about myself as 'broken in a sometimes useful way'. My burden in life is to manage and control the characteristics where they are disadvantaging me, and position myself as much as I can in positions where these characteristics are useful assets. To hook into your winter metaphore: If you always feel cold, working in a tropic resort where others would feel continually hot, sweaty and uncomfortable might just be perfect for you.
Where it crosses the line for me is neurotypical people saying 'oh I don't need winter clothes, and anyone who does is must be to use a shortcut and should just jog themselves warm instead'. Or even ADHD people saying: 'My feeling cold all the time is a superpower, you should just go work in your local sauna and you'll have no problem at all!'. Great, happy that you guys don't have this issue or figured out a workaround, but I do, and I need my damn winter clothes. I wish I didn't, I'll look for warmer situations, but I need them right now.
Apologies for drawing that conclusion, but usually the argument presented is something like "amphetamines / stimulants bad especially in healthy adults -> usage on the rise due to increases ADHD diagnoses -> diagnoses potentially fraudulent? -> pharmaceutical companies are incentivized to sell stimulants -> ban stimulants entirely before it gets out of hand and rejoice our win on late stage capitalism and druggies".
I might've misunderstood the point you were trying to make, but saying 'society causes the issue' usually is followed by 'therefore treating with meds is silly and we shouldn't do that'. The latter part is what I take issue with, not the former, and if you weren't advocating for that then we have no issue and are in agreement.
I'd love to find a constructive way to change society for the better so ADHD is not as much of an issue, but personally, I don't see that happening. And I do see anti-intellectualism and puritanism on the rise, and with it, a movement that wants to take my crutches away to deal with society as it is. Maybe you weren't one of them, but I saw you making similar arguments, which is what prompted me to respond.
Calling the group of all stimulant medications "speed" is not a great sign either if I have to judge where somebody stands in this debate btw. Dextroamphetamine, one of the ingredients in Adderall, is similar to other amphetamines as found in speed, but for instance methylphenidate aka Ritalin only partially has the same working mechanisms, blocking the reuptake of dopamine and norepiphrine but not helping the extra release of more of these neurotransmitters. And that comes with a different profile in terms of addiction risks and whatnot. It is not helpful to call all of these medications 'speed' as if all ADHD people are buying potentially contaminated stuff on street corners from shady dealers, manufacturing this breaking bad style in their shed. Managing ADHD is done with clean, medical-grade, typically less-potent versions, with tightly controlled doses, closely managed by a licensed medical professional. Framing ADHD patients taking their prescription medication as speed users is not helpful. Even if you don't have any ill intent, reinforcing the belief we are all essentially the same as junkies just provides ammo for people who do want to take this away.
Even if you were 100% correct and the world is broken, fully causing ADHD as a disorder: Please fix the world FIRST and only once proven ADHD is caused by what once was the shape of western society and no longer applies, THEN you get to take the metaphorical wheelchair away.
The alternative is that you prevent millions of people in managing their disability while asking them to bet on your view of the world AND on our collective ability to change it. In the best case scenario where we manage that shift, that's what, 10 years of my life gone while society adjusts? Will you write my kids a nice letter explaining them their dad is going to be a deadbeat the next 10 years while we fix society, because somebody on the internet thinks daddy shouldn't be on stimulant medication?
You're just not presenting an attractive deal to anyone, whilst very politely telling disabled people making the best of their shit situation that their crutches should not exist. Hell, maybe they shouldn't need to exist, but how is that my fault? And while I can't tell if you stand on the side of 'using meds to manage ADHD is a failure of self discipline and morality', but if you do: I promise you most people with ADHD have more self discipline in their little toe than others do in their entire body. But self discipline doesn't make a cripple walk, as much as it doesn't make my brain make the chemicals I need to put my body into action. I've spent enough time of my life flogging myself into action, believing I was a fundamentally lazy human, I'll take the meds.
I suppose that's true, in much the same way that chemotherapy treats the symptom of cells spontaneously deciding to replicate in your body. That does not mean we judge people battling a cancer diagnosis and tell them to pursue non-medicated approaches because it's "just treating symptoms".
If you encounter a bit of bitterness from the ADHD community online, let me provide some perspective: I have been called lazy my entire life, I have wondered why everybody could just do stuff and apply themselves. Why couldn't I just clean my house, do my homework, keep on top of chores, or even find the energy to play games after a day of work? I only found out as an adult I have a disability which makes all of that an uphill battle for me, INCLUDING finding the motivation for the fun stuff. There is an easy fix for this, some meds that take care of SOME of the problem. They don't fix it in much the same way that a wheelchair doesn't fix the legs of a crippled person, but it sure is like playing life on easy mode if you're used to dragging yourself around by your arms. And now I'm stuck explaining this to people who have done the barest minimum of research and who say 'oh it only is treating symptoms'. They have the audacity of calling me lazy (again!) for not training my arms more to overcome my disability that way. And my response is simple: You can take my metaphorical wheelchair over my dead body, and if you were in my position you would feel exactly the same way.
To provide a European/Dutch perspective: I’m pretty sure that as a small employer myself, I am very much disallowed from using those mechanisms to actually inspect what employees are doing. Automated threat/virus scanning may be a legal gray zone, but monitoring-by-default is very much illegal, and there have been plenty of court cases about this. It is treated similarly to logging and reading all email, Slack messages, constantly screenrecording, or putting security cameras aimed at employees all day long. There may be exceptions for if specific fraud or abuse is suspected, but burden of proof is on the employer and just monitoring everyone is not justifiable even when working with sensitive data or goods.
So to echo a sister comment: while sadly it is common in some jurisdictions, it is definitely not normal.
Minor pedantic correction: 2.5gbit, 5gbit and 10gbit RJ45 is getting more affordable and more common, and for short runs should run over CAT 6 and CAT 6a fine, and plenty of reports it does ok on short runs even on CAT 5e. With devices like the USW Flex Mini 2.5 at ~50-60 EUR / USD, you can affordably outfit your home for higher than gigabit speeds without rewiring everything with new CAT cable or fiber.
Over here in NL we now get more and more access to >1gbps speeds, the office of my small business for instance has a 4gbps connection, and the ISP offers up to 8gbps on a standard consumer / small business package. We're in the process of upgrading our gear to take advantage of that. With WiFi 7 we've seen some real world throughput speeds of 1800-2000mbps going through a Ubiquiti U7 Pro straight to the ISP supplied router.
I wasn't really keeping up with networking gear, so I was pleasantly surprised when I looked into this stuff recently and figured out the gear has just magically gotten better and running 2.5gbit everywhere is surprisingly easy.
I've wondered this as well, I'd love to hear from the mods as to how many false positives vs true positives this generates. Us, the lowly users, only spot it when it mangles a title, but does it actually provide some tangible benefit?
I don't want to judge this 'feature' too harshly without that data, but couldn't 80% of the value of this be achieved by putting the text 'please don't editorialize titles of submissions except to de-clickbaitify them' in the submission form?
Shipping has always been a strong point for Ikea. Just look at how efficient most of their stuff is packaged, usually into very neat rectangular packages. They can fit a ton of stuff into a small volume, which improves shipping cost a lot. Last mile will always be relatively expensive, but shipping a truck load or a 40ft shipping container is relatively cheap. In the end it all comes down to cost per unit of volume, compared to the raw product cost per volume. A dense product will not nearly be affected by high shipping prices as much, and due to Ikea being insanely good at packaging, I expect they are outcompeting the rest of the market at this game by a decent margin.
Same, I really have nothing bad to say about the Tailwind design system.
I suspect there is a lot of halo effect going on, where if people were asked to judge the pure code aspect of Tailwind vs the design system separately, they'd be much more negative on the code aspect of it, but because the two are conflated and one is very good, Tailwind as a whole is judged to be much better than I personally think it deserved to be. It doesn't help either that forming an opinion on an architecture of code is difficult, and forming an opinion on a designed web page in front of you is easy.
My personal poison is just defining the Tailwind color system and a very basic system of sizes in CSS variables and then using that all over the place. It leaves an escape hatch in place of just using 5px when you really need to, but most of the time we use var(--col-grey-200) and var(--u-4) or var(--u-8)
I'm not necessarily arguing in favor of deep cascading, but I don't see how something like this:
.form-element,
.button {
height: 32px;
}
Is something that can be easily achieved with Tailwind without either using @apply all over the place, effectively now doing regular old CSS but with Tailwind syntax, or by using JS/TS variables extensively making the styling pretty hard to read. Either way, I'm not saying Tailwind makes it impossible, but it sure doesn't make it easier. And while Tailwind has a bunch of benefits, especially where it becomes a design system rather than a syntax for CSS, you can achieve ~90% of that by just defining a bunch of color and size variables at the root of your CSS and using that.
I don't mind Tailwind too much for 1:1 mapping with DOM elements, but I also don't really see why inline styles for that case would be bad.
Starting from 'what looks good' is putting the cart before the horse. Making a UI usable and well laid out first is key. Practical UI and Refactoring UI are great resources, as long as you read them through a lens of 'what works well?' instead of 'what looks pretty?'. The author is absolutely right in that alignment and consistency are important, but that should really be your starting point.
Building a good user interface is fundamentally an engineering challenge. I see roughly two camps in building UIs, one designing a pretty picture and then tweaking the CSS until it looks like the picture, the other treating the CSS as rules of how the UI should behave. A simple example would be using display: flex; gap: 32px; on a parent of two elements instead of margin-right: 32px; on the left-most element. While the end result is identical, specifying the gap on the parent is better, because it puts the responsibility for spacing in the correct place. This also goes for the way you define CSS classes and rules, if two values are linked, like the height of your buttons and the height of your input fields, then try and capture that in a single rule, or extract it out to a variable.
A lot of building good UIs becomes much easier once you adopt the engineering approach. Consistency is almost built-in at that point, and that automatically makes your UIs better and easier to understand. It keeps your CSS more maintainable too.
While I'm sure there are ways to achieve this with Tailwind, generally I tend to see developers do the exact opposite when they use tools like that: just define everything with atomic classes inline, and forget all about the relations of styling rules to eachother. Tailwind has some great concepts, like defining a limited set of values to choose from, but be careful to keep the engineering, rules based way of building UIs.
There are so many times we've gone a direction in our products only to figure out that while we could make the page look pretty, it never would work well. It always ends up being some version of 'if we go direction X, then features Y and Z will have to be shoe-horned in and it'll look ugly'. When you get that feeling, take a step back, come up with some different approaches, and go with a better one.
The "make it pretty"-step should really be the last thing you do. If you design your UI with heavily visually simplified components and in black and white, it should still work and feel right. Make it work right, and the pretty will come.
That still really isn't a reason. I think I can extrapolate where they are going with this, which will be something like 'I have a ton of 720x400 hardware CRTs sitting around that I need to use/support/deal with'. But that is never explicitly stated, you can completely read it as 'oh it's neat that 80x25 matches up with the number of image lines on old CRT displays and here is the math to show it'.
Now if Fable had an easy jailbreak like this that allowed you to attack remote targets that'd be a different story but I genuinely cannot see how neutering its abilities to 'fix' code you already have access to is sensible. It would destroy the value of the model. And don't forget, any actor not abiding by the same rules could develop an model for offensive use just fine, so this protects you against exactly nothing but does destroy a potential defense.
In the end this all comes down to legislation, in much the same way platforms are not responsible for copyright violations IF they abide by some rules, the same has to happen for AI providers. If you have a process for reporting 'jailbreaks' on illegal actions, and prevent users doing illegal stuff on a best effort basis, the rest of it should really just be individual responsibility. If a user wants to use an LLM to crack systems, fine, that's already illegal.
If Tesla FSD deliberately hit somebody, holding Tesla liable is fine. If you messed with FSD until you finally got it to hit a person, then you should be liable. Outlawing FSD because it could theoretically be tampered with is just an odd stance imho.