I hope they don't. These are large language models, not true intelligence, rewriting a soul.md is more likely just to cause these things to go off the rails more than they already do
They aren't truly intelligent so we shouldn't consider them to be. They're a system that, for a given stream of input tokens predicts the most likely next output token. The fact that their training dataset is so big makes them very good at predicting the next token in all sorts of contexts (that it has training data for anyway), but that's not the same as "thinking". And that's why they get so bizarelly of the rails if your input context is some wild prompt that has them play acting
Well, the way the language is composed reads heavily like an LLM (honestly it sounds a lot like ChatGPT), so while I think a human puppeteer is plausible to a degree I think they must have used LLMs to write the posts.
I hate when people say this. SOME engineers didn't care, a lot of us did. There's a lot of "engineers getting a taste of their own medicine" sentiment going around when most of us just like an intellectual job where we get to build stuff. The "disrupt everything no matter the consequences" psychos have always been a minority and I think a lot of devs are sick of those people.
Also 10x salary?! Apparently I missed the gravy train. I think you're throwing a big class of people under the bus because of your perception of a non representative sample
I'm guessing this was probably accidental/weird consequence, but it does raise a much scarier possibility. If someone wanted to set AI models out against people as a reputational attack dog (automating all sorts of vicious things like deep fakes and malicious rumors across sockpuppet accounts..) I mean, are there really any significant obstacles or ways to fight back? Right now slop is (mostly) impersonal, but you could easily imagine focussed slop that's done so persistently that it's nearly it's nearly impossible to stop. Obsessive stalker types have a pretty creepy weapon now.
Pretty sure it's illegal. You can get peptides (even GLP-1's) legally from various sources for "Research Purposes", but they're marked as "Not For Human Consumption" (even though, on the sly, I'm sure they're aware people are buying these for human consumption and they provide purity tests etc.) What makes it illegal though, is when you say it is for human consumption, or worse market it as a treatment for a disease. That's when you actually need FDA approval.
Hard disagree on this. $26.50 sounds like a nightmare 10 years ago, let alone now. There's a lot of places in the US where having a car is essentially mandatory (actually, most places). If you can't afford a car, that limits where you can live to mostly urban areas, which then pushes the housing cost up.. and by the way, housing costs are always going up, and no, you won't be able to invest in a home, you've been priced out by developers and speculators.
Not to mention you need to be able to save money for unemployment and rainy days..
I think you're ignoring how much poor people rely on each other and relatives to get by. That's our societies "safety net". That doesn't mean they're "thriving" or even comfortable, nor is it even sustainable (what happens when mom/pop die or require assistance and can't help their kids anymore?).
9000/yr for a car alone isn't crazy at all, just look at average car prices. I just had to do my vehicle renewal today and it was $500 for a 5 year old car that's not particularly expensive! If I look at insurance and car payments, I easily spend over 700 a month. This is on a 30k car, so it's not like I went and bought the biggest luxury vehicle possible.
The flip side is that the people being relied upon are performing uncompensated labor or providing other unpaid services, which is not a healthy state of things. This very dynamic can end up trapping these people in poverty and hinder their access to more productive arrangements.
My feeling is technological unemployment could go through the roof the coming decade. First, most everyone will be beaten on cognition and then also on real-world tasks by versatile robots.
Eventually there has to be a mechanism that continues providing liquidity for nothing or...
The average total cost of car ownership in the US in 2025 was about $12,000. $9,000 is already a huge underestimate of what the average person is paying.
I lived in a blizzard ridden area using just a 250cc motorcycle, year round, including riding it on the interstate. Layer enough layers, use heated gloves, etc you can easily get by with just a ninja 250, you're not going to burn more than $3-4k a year on that no matter hard you try.
You don't actually need a car unless you have a child or a tradesmen with tools or something like that, a small displacement motorcycle will still take you to 99.9% of the jobs in the lower 48.
Leaky abstractions have always been a problem. Sometimes people like to use them as an example of "see, you didn't understand the assembly, so why do you care about... X". The logic seems to be "see, almost all your abstractions are leaky, why do you care that you don't understand what's happening?"
A few comments on that. First off, the best programmers I've worked with recognized when their abstractions were leaky, and made efforts to understand the thing that was being abstracted. That's a huge part of what made them good! I have worked with programmers that looked at the disassembly, and cared about it. Not everyone needs to do that, but acting like it's a completely pointless exercise does not track with reality.
The other thing I've noticed personally for myself is my biggest growth as a programmer has almost aways come from moving down the stack and understanding things at a lower level, not moving up the stack. Even though I rarely use it, learning assembler was VERY important for my development as a programmer, it helped me understand decisions made in the design of C for instance. I also learned VHDL to program FPGAs and took an embedded systems course that talked about building logic out of NAND gates. I had to write a game for an FPGA in C that had to use a wonky VGA driver that had to treat an 800x600 screen as a series of tiles because there wasn't nearly enough RAM to store that framebuffer. None of this is something I use daily, some of it I may never use again, but it shaped how I think and work with computers. In my experience, the guys that only focus on the highest levels of abstractions because the rest of the stuff "doesn't matter" easily get themselves stuck in corners they can't get out of.
reply