HN user

rekabis

624 karma
Posts1
Comments839
View on HN

“Even if the motivations were ethical, this is clearly an example of AI out of control,” lead researcher Aengus Lynch told The Bureau of Investigative Journalism.

For the first time ever I am frightened more by the company than by the AI.

Up until now I only saw the company as gangs of madmen looking to herd cats. Now it’s clear that they are less sane than their wildly hallucinating products.

As of the current time, our quantum computers are utilizing a hilariously low number of qBits, which severely limit the number of alternative universes that can be leveraged for calculations. As the number of qBits in use increases, the processing power expands appropriately.

As another commenter has said: it’s not if, it’s when.

The single best signal for identifying a phishing domain is the domain registration date

I guess the authors have never heard of domain washing?

Because IME a majority of domains from serious and large-scale spammers are not new - they are abandoned domains, where the spammer is counting on that domain’s prior reputation to bypass any filtering.

I was slowly ramping up to run with them as a provider on the personal routers of my entire family. Mainly because it would have saved me the headache of setting up my own VPN provider.

Shame, that. They seemed so promising.

Screw the oil crisis, it’s going to cause a food crisis over the next 18 months.

Modern industrialized agriculture absolutely depends on manufactured fertilizer. It’s gotten to the point where demand is almost perfectly inelastic - farmers literally cannot grow a successful crop without it, and in the amounts needed. They cannot just “tighten the belt” on fertilizer usage - fall beneath a certain level, and the entire crop becomes unsellable.

And yet, a good 60% of all Urea - a critical fertilizer base component - comes out of the gulf states via the Strait of Hormuz.

We are going to see the opening scenes of this crisis over the winter, as the southern hemisphere countries start reacting to fertilizer price spikes as supply gets massively constrained.

And next year through the fall time will be the major acts, as northern-hemisphere crops cannot be grown.

Remember: it takes time for products to be manufactured and move through supply chains. The southern hemisphere countries will be in the process of fertilizing now through the fall, depending on crop. Northern hemisphere countries will be fertilizing from late fall through next summer. We could easily be looking at severe shortages of fertilizer lasting into 2027, leading to a catastrophic yield collapse of agriculture for two years running.

And this can and will also impact meat production - every species that ends up on our dinner plate is also supported by industrialized agriculture, as that provides the food that it eats.

I really hope y’all all have functional vegetable gardens to fall back on.

Meanwhile, in my corner of the planet we have an ISP called Telus, who provides both IPv4 and IPv6 for their home users, but has only IPv4 for their business users.

Exact same fibre optic lines. Exact same fibre modems. Exact same Dmarc points. Exact same networks.

But business users, who frequently run servers on their connections, don’t get IPv6.

Make this make sense.

This add-on was vibe coded using GLM5.2 and built to solve the needs of my team. I was hesitant to share it, as I generally do not like vibe-coded projects. However, given that a tool like this is sorely lacking from the ecosystem, I decided to share it anyway.

Thank you for your honesty.

If you are wanting a laptop for general or creative work, I would absolutely agree with you.

If you are wanting a laptop for business work with Windows-only apps that are niche enough to never get support from WINE, I would honestly recommend Dell business-class laptops.

The downside is the price: these are built like tanks, somewhere between consumer and toughbook standards, and as such, are commensurately expensive. The ones I am talking about are the “pro precision workstations”.

Before the AI price spikes (mid 2025), it was not unusual to spend $2k CAD on a midrange model, $8k CAD with all the bells and whistles. But as a result, running one for 6-8 years is not uncommon.

And since last fall, it seems that prices have jumped by $2k CAD on the midrange machines to as much as $14k CAD on the high end thanks to RAM, SSD, and GPU prices. Like JFC, plumping out to 128Gb is now +$8k CAD all by itself, and +$3k CAD just for a decent SSD. The top end machine went from something around $8k to $22k CAD, and even dropped the options for 256Gb and 512Gb RAM.

MSSQL Server has had this capability since forever.

You can even graphically drag-and-drop column order from within SSMS (SQL Server Management Studio), although it’s always seemed sketch AF to me, and I wouldn’t want to do this with a large table unless I had backups and the site/app was set offline due to maintenance.

Even Entity Framework (DotNet) has the ability to define column order, but only on table create (for sure). I cannot remember if an EF migration can insert a column anywhere into the middle or not.

Around the turn of the century, my years of web dev work started taking a toll on my carpal tunnel. Some research led me to track balls (Kensington Expert Mouse) and the DVORAK layout. I managed to convert quickly enough to never have a truly bad flare up, and to this day I am rocking that layout.

Pro tip: once you have committed, you can get a hard-coded DVORAK keyboard from UNICOMP, who make modern versions of the IBM Model M buckling-spring keyboard (which is often called a “desk yacht” due to its massive size). This allows you to have DVORAK without having to mess with your system’s default keyboard, and allow you to type on other people’s systems by just dragging your keyboard along.

You need to add a customization to your order and specify DVORAK, but the keyboards are absolutely spectacular and last for many years. Be sure to also pick up the keyboard protector if you are not keen on breaking it down twice a year just to clean it. The plastic protector is low-end lame, as it is not exactly form-fitting (vacuum-moulded instead of form moulded), but it is better than nothing.

Why? The vast majority of people will never be able to afford a luxury apartment.

Now affordable apartments, which have dispensed with the limestone or quartz countertops and the imported Italian tile floors and all the other fancy bits and baubles that jack up the price by 50% or more, THAT is what we need a lot more of. Not drab or dreary accommodations, but thoughtfully designed places that maximize livability while minimizing cost.

Let me see a low-cost but functionally robust and well-built apartment that won’t need major repairs or upgrades for 30-50 years, and I would gladly see a small part of my taxes carved out to lower the monthly costs that renters are faced with.

The Vienna model is particularly apropos to this situation.

Because what we need is housing that is trivially affordable to ALL of the working class, and not just the parasite class.

And the true measure of any society is how they support and take care of their most vulnerable members.

I feel this is important, because the vast majority of attention to date has been on the harmful effects of how students use AI themselves, to game the system or even outright cheat.

From the paper:

Teachers are rapidly adopting generative AI, yet there is little causal evidence on whether AI-assisted teaching improves or harms student outcomes. We conducted a randomized field experiment across a chain of middle and high schools in Turkey, assigning teachers to receive a generative AI teaching support tool or business-as-usual control. The majority of teachers used AI for lecture material, syllabus, and exercise creation.

I disagree. Testing with multiple rounds of feedback and iteration, with merging and related work done, with everything in production and available to the users is done, but is it perfect?

Not necessarily.

I see it as having dealt with all low-hanging fruit. Of handling anything reasonably expected at that point. Bugs will remain. Exploits will doubtless be found. But so long as all reasonable efforts were made to find and deal with them, that is good enough.

Because a perfect product will never be done.

And a done product cannot possibly ever be perfect.

For me, I take a page out of IBM’s book:

A computer can never be held accountable, therefore a computer must never make a management decision.

Except this should really be “any kind of choice or decision” in general, not just management ones.

This means that AI should never be allowed to make any kind of decision or be able to employ any kind of choice, including what code to create and how the program gets constructed, because it can never be held accountable.

And capitalism demands the “business as usual” path that is also the worst-outcome-possible path.

Wholesale civilizational collapse has already been modelled to occur between 2035 and 2045 to an extremely high degree of confidence, with a 50-80% collapse in human population to occur before the end of the century. You can’t maintain a post-iron-age civilization through that degree of collapse.

I would even posit that a post-bronze-age civilization would be difficult enough to maintain. Because you need a fair amount of tech to work iron on scales larger than a blacksmith’s shop.

And these models do not include major attributes of climate change, such as wet bulb temperatures forcing billions to migrate into first-world countries to survive, or how regional ecosystem collapses cause trophic cascades that ripple around the globe. They only looked at simple things like desertification, topsoil loss, chaotic weather (unreliable rainfall) and so forth.

If it works, it works, and you could be directly helping thousands of other small businesses that will never compete with you.

That’s awesome.

Eeehhhhhh… while I am cautiously optimistic of AI, everything I have seen to date still makes me want to avoid its creation of code. It still hallucinates far too wildly, over engineers far too eagerly, and can only crib from prior examples; it cannot look at a situation and use judgement and nuance to create truly appropriate and effective code.

So using AI as a guide while you write code, sure. It’s probably the best rubber ducky ever built. But letting it actually write the code? Nah, man. That’s the path to unmaintainable slop crammed full of technical debt.

what's one piece of advice you wish someone had given you during your first year?

Cross-train in another field. Literally anything outside of CS.

Employers are doubling down on not hiring juniors, especially GenZ, despite the AI pullback. Even four years from now, you will likely be entering the worst job market that the CS field has ever had for new entrants.

My condolences, but that’s how things are. It sucks and it’s moronic but employers can’t handle people who are resistant to abuse and exploitation.

Desertification, wet bulb temperatures, heat stress, loss of food in the environment, ecosystem collapse due to climate change happening too fast for ecosystems to migrate, and the massively increased range required to find food causing them to not find mates.

Point of contention:

It can no longer be denied: Climate change is progressing unabated and accelerating. In 2023 and 2024, global average temperatures were 1.5 degrees Celsius above pre-industrial levels for the first time. It is even possible that the 1.5-degree limit for global warming agreed upon in Paris may already have been permanently exceeded.

The value they are talking about is the political definition, and not the scientific definition. The political definition requires a +1.5℃ increase to be unbroken across a number of years - 20 years, as set by the IPCC - before it is “accepted”.

Not the first year in which warming never again drops beneath +1.5℃, but 20 unbroken years of said minimum.

Many scientists consider +1.5℃ to have been reliably and consistently breached at some point in the very early 2010s.

And by that same metric we might have already blown past +2℃, which may have happened by 2023.

The reason why this is so alarming is because +4℃ is the point at which megafauna - all animals over 45kg, like humans - can no longer survive on this planet. They go extinct.

There are adults alive today which will see this come to pass.

I cautiously agree with this sentiment, in particular the need for any corporation to have trust as its foundation. Without trust, only scammers will leverage the corp’s product or service. Trust is vital to attracting any kind of legitimate and long-term client, from individual consumers to large and slow-moving behemoths of business.

And AI cannot generate trust. It is incapable of doing so.

Its wild hallucinations and overconfident falsehoods rub me the wrong way.

I am cautiously optimistic about AI, but not only does it cheapen the dev experience and suck all my enjoyment out of coding, but it also is far too unreliable at this point. It’s a swarm of ADHD juniors with goldfish memories and hopped up on acid and magic mushrooms. It takes more energy and effort for me to get it to produce minimally acceptable output than the energy and effort I need to produce superior output entirely on my own. And my objective is not slop at high speed. It’s quality work at lower effort. That’s something that is beyond the capabilities of AI at this point.