kypro

UK is crazy for this. Border police can detain anyone they like for literally no reason and if they do you have no right to lawyer or no right to silence. Your devices can also be taken and you must hand over any passwords.

I know the UK is an authoritarian state and not really comparable to the US, but as a Brit it's kinda weird hearing Americans complain about flock cameras and arbitrary border searches – these are common place things in the UK and no one cares (well, apart from me and maybe one or two others).

Quick point on capitalism – the profit seeking motive is what makes the economy efficient and ultimately is what results in higher wages and living standards for everyone. We should be careful not to over-index on the direct effects of the "profit motive" such as profits to asset holders, while overlooking the far more important downstream benefits to society and workers. It's not win-lose.

But yes, it seems fairly obvious to me that if AI will effectively displaces all human economic value then the economy will not work for humans.

The only point I'll make on that is that I suspect you're worrying about the wrong thing. In a world where humans are economically worthless they're also politically powerless and entirely dependant on the decision making of the more intelligent artificial beings they've created. In this world your employment income will quickly become the least of your concerns.

Surely this is way too broad a statement?

There are small A/B tests which absolutely make sense. You often see marketing sites making small tweaks to banners and copy. It's not that one has worse UX or even that one is objectively worse, just that different users have different preferences and it's often difficult to know exactly what will work best.

Similarly you can be very confident of something, but A/B testing it still reduces risk. Any significant change should probably always be rolled out to a small fraction of the user base first in case you accidentally change something for the worse.

I agree if you're talking about some BS experiment where a company uses A/B testing as an alternative to putting the hours into product design and user research.

Surprised myself a bit with the pillar post box being my winner.

I love the British phone box too, but I think because my childhood neighbourhood was built in the 80s I never really saw the traditional red phone boxes when I was young. They were all these, https://en.wikipedia.org/wiki/KX_telephone_boxes

The post boxes are lovely though and remind me of my England more than any of the others.

I'm normally happy to jump onboard the HN hate train, but I think this time it's more that the launch was full of marketing slop about AGI and positioning Jev as an LLM competitor, when as you note its basically just a general purpose classifier.

This isn't to knock the team at all, but imo the most impressive thing about Jev is that no one did it before. The second most impressive thing is that there was such a large number of people that didn't realise that classifiers were useful.

> I stand by what I wrote earlier this year: Alcohol was a net positive for me. It helped influence the bond of friendships, hundreds of wonderful nights in beautiful places, and in many ways, the founding of FireHydrant.

I feel similarly about alcohol... I came very close to drinking myself to death after a decade or so of heavy drinking, but I think it was also the case that alcohol probably saved me and may have been net-positive overall.

I was very depressed at that time and when things were bad it was useful to know that soon I would be drunk and the pain would go away. That's why I drank.

Now every day is gray and there is no relief and probably will never be. I'm coping and will continue to coop for the little time we have left, but I miss alcohol.

Maybe that's how all alcoholics feel.

It was also very good during the pandemic. Google results at the time were extremely biased towards Western government narratives. Some mainstream articles which took a skeptical view on lockdowns and the virus orgin seemingly would not show regardless of how you tried to search them on Google.

Yandax on the other hand tended to be more neutral.

I deliberately used non-emotive language and didn't explicitly hint at how I see things evolving with AI – otherwise what's the point if it's just telling me what I already know?

Instead I tried to stick to my view on both current and future capabilities of AI, and how it's likely to progress in the coming years.

My result was closed to Hinton which isn't that reflective of my actual view (I'd put doom at around 99% at this point). But that's probably just because of how I answered.

I like it though! Especially how it clearly exposes how unreasonable Ed Zitron is, and perhaps myself =)

> Undoubtedly, a transition of this magnitude could be enormously disruptive. But there is another possibility worth considering: that the extraordinary speed of progress itself compresses the period of disruption.

> To me and to our team at Infinite Ascent, this future of expanding abundance driven by an explosion in intelligence is something we deeply believe in, and something toward which we have chosen to direct our resources.

> But our underlying belief is one of profound optimism: that greater intelligence can ultimately produce a world of greater abundance, greater knowledge, and greater possibility, while allowing us to build in a way that benefits both intelligent life and the natural world around us.

I typed out a much longer comment going over some of my extreme disagreements with this article, but it's all so pointless. I'm just yelling at the clouds.

I just hope when this doesn't work out as planned these people reflect for a minute on their contribution to the suffering they have caused.

I guess I'll just remind people that thanks to technology we already live in a world with an abundance of food. What we don't live in is a world in which people would rather save a hundred starving kids over taking their family to Disney Land for a week. And so people die in completely avoidable ways every day, and we don't even lose sleep over it. That is why this article is wrong. If you make people economically worthless in a matter of years, I assure you, that will not end well. And that certainly isn't the only risk of ASI.

Apple is a private company. They don't have to comply with these demands if they don't agree with them.

I'm a Brit and I've advocated that Apple pull out of the UK over this. It annoys me that these companies seemingly have no principles and will just comply with our government's authoritarian demands. As a citizen there's nothing I can do, so it's really annoying that those who could be doing something choose to do nothing for a little extra profit.

> There is, if there were no advantage a whole lot more species would die immediately after reproduction like the octopus.

I didn't just meant extending lifespan, I meant extending the time an individual is fertile.

> The utility of genetic diversity puts a lot of pressure against this.

High levels of elitist selection lowers diversity and can have a negative impact on a population's adaptive ability, but there is a sweet spot to diversity. Maximally diverse is just noise after all.

A small amount of elitism is generally what you would expect to be selected for, and we certainly see this other forms in most creatures, but never in the form of extended life span. That makes me suspect there's probably something fundamentally difficult about extending an organism's lifespan given our biological architecture and that's why evolution has favoured other forms of elitist selection.

As others have noted it's obviously more productive to use AI to write most of the code and for you to focus on the areas where you can add value.

However soon you will find have no real advantage over AI in any task. Therefore caring about the skills which are of fading value today makes almost as little sense as caring about the skills that mattered yesterday.

> What is gonna be sought after in the coming years: someone that persevered with manual coding and debugging but is slow and uncomfortable around agents, or someone that has little coding and debugging skills left but knows how to orchestrate dozens of agents effectively?

Neither. If your concern is your career or income then you should have a real plan based on how the world is realistically going to look in the future.

> In a world where anyone can create almost anything, having good taste could become increasingly important.

The latest models from OpenAI and Anthropic are starting to develop taste imo. Specifically Astra and Opus 5.5.

It's not quite there yet, but also far from the slop models were producing this time last year.

You could build something really useful, but if I can replicate 95% of that in a weekend with Claude I'm obviously not going to pay you for it.

Some SaaS products have network effects, but that doesn't appy to new software.

Starting a SaaS business in 2026 seems a bit silly to me. Not saying people won't make money, but you'd have to be pretty lucky.

My guess is that this is half the story. It seems like there would be some advantage to some creatures of not aging, or having individuals who have significant longer life spans.

This is common in evolutionary algorithms anyway so it's notable we don't see it in nature, at least not in the form of extended lifespan perhaps measured by average lifetime heart beats. It's easy to imagine a world where nature might want to keep the alpha lion around longer than the others so it can have more offspring, for example.

Does this work for foreign (non-state) actors using Chinese open source models? That's going to be the larger problem.

I think he is wrong in regards to studying computer science assuming the objective is to do something that leads to success in a future world where AI is writing 99.9% of code.

Obviously there is some value in learning just about anything, but that isn't actually how people decide what to study in practice.

> On one hand you have AI enthusiasts who are shipping more than they ever have, they are super excited about being able to use the models and build everything they've ever wanted to. Utopia, amazing.

You can't ship faster than you can comprehend, unless you stop caring about comprehension.

Up until the last year the best developers were generally those who valued comprehension and wanted to really understand the libraries, code and infrastructure of their projects because that comprehension was what led to quality software.

Those who didn't care tended to ship buggy crap and weren't great devs.

In my experience those who didn't really care about deeply understanding the systems they were working on are loving AI. But those who did care are now literally unable comprehend the state of the codebase and are being forced to work in a way that's extremely unnatural to them.

I say this because I've personally struggled with this over the last year as I've was one of those people who spent a lot of time building comprehensive mental models for the systems I worked on. Now I just assume the AI understands and just do what I need to do to get whatever functionality I'm working on over the line. I don't like it and I've lost all technical interest in coding at this point, but it is what it is ultimately. Coding is no longer about comprehension, it's just about asking Claude to do something.

If there's one thing that I attribute the relatively modest success I've had in my life it's that I'm consistently predicting where we'll be in 10 years and optimising my life to position myself well when that comes.

Right now I'm strongly advising people not to care too much about narrow academic or intellectual pursuits because I think those skills will be almost entirely redundant to AI in a few years.

Learning to code today seems absolutely insane to me honestly. If you just want to understand at a high level how code works, then cool spend a week or two on it, but obviously outside of very niche edge cases which you can't build a career around humans won't be reading and writing code for much longer.

I'd even expand on this to say people should broadly avoid tech as a sector to learn anything in at this point. There will be a lot of disruption in the coming years. Hardware and software are going to change to better suite an AI-first world and this means almost any tech skill run a risk of disruption.

That said, I struggle predict past 10 years these days. In some ways we're now just weighing different sized infinities. In a world of AI and robotics perhaps the only good bet to make is that human are of no value anymore. This is the world I am personally preparing for. Any skills you learn today would ideally be in preparation for this.

> Since AI is trained on "the internet" it sees and learns a lot of bad habits, poor coding examples, shortcuts, oversimplified solutions, and therefore, by default, produces code that can be improved.

This is a huge simplification. AI is also trained on tweets, spam and explicit content, but ChatGPT isn't constantly swearing at you to send your bitcoin to a Nigerian prince.

The mistake people make here is that just because something is represented in the weights, doesn't mean it's representative of its output. And the reason for that is largely the same reason for why if I show you lots of bad code you don't become a worse coder. If anything you'll see bad code and learn to avoid writing similar code yourself.

So long as quality is being reinforced and that process is refined and improved then "model collapse theory" and arguments about how AI will write bad code because it reads bad code will continue to prove incorrect.

> Having an intelligent machine doesn’t mean you can somehow mind control individuals.

Excluding advanced forms of psychological manipulation, sophisticated neurological drugs and neural simulation for moment.

Are you suggesting that it's physically impossible to create a device which could induce electromagnetic currents in the brain which could in-theory either effectively control, or greatly influence someone's decision making?

I understand it would be extremely hard for humans, but can you explain why you're so confident that this would be such a hard problem that even an ASI couldn't solve it?

> That won’t happen.

Why?

> But also, whatever benefits are unlocked will be owned mostly by a small group of individuals, definitely not available to humanity as a whole.

It's unclear to me if this is the good scenario or bad scenario. This would be good in your view right? At least I hope you're right.

All technology on the tech tree which requires intelligence to unlock will soon be available to humanity – mind control, population exterminating bioweapons, new ultra destructive kinetic weaponry, perhaps even a cure for cancer.

I don't think it's that. Most people simply don't care.

I'd stop paying them, but it always amazes me just how unbothered the average person is to watch 3 minutes of ads for 10 mins of content. Especially when it comes to platforms like YouTube.

> My gut is telling me ECS is the better choice for this team - simpler, no operational maintenance, almost no additional abstractions beyond the AWS primitives people will need to learn anyway. A consulting team hired before I was brought in are pushing EKS, but I’m not convinced that the improved devex is worth the increase in complexity and operational burden for our team.

Yes, I'd say 100% ECS given what you've said here. Don't add complexity until you need it. When you have a larger and more mature devops team you can always reassess if it's time to consider EKS.

Also, definitely have multiple AWS accounts either way. This is very standard and arguably best practice. AWS have management accounts specifically for this reason. Keep developers off prod and on a AWS separate account where you (and they) can be 100% certain they can't do any harm.

But at the end of the day, neither are strictly wrong and what you're asking here is largely a question of corporate politics. If you don't feel you have a strong voice yet, then just try to have a good sense of the pros and cons of the different options you're considering and present a well reasoned opinion. If you're told you're wrong or you're overruled, then that's okay! Ultimately either will work, and right now it's probably better that you focus on building trust before being too opinionated.

I can only speak for myself, but I definitely don't see my inability to socially align with others as a superiority thing. It can give me some predictive advantage when there's social stigma around a certain belief. But that's really about it. In most cases, especially professionally, it's very hard to navigate and I'd rather be naturally more aligned with people.

But yes, my profile is cringe and I'm sorry someone has downvoted you for sharing your view. There should be less of a taboo about telling people we dislike them if that is how we feel. It's better than to lie.

Do you have an argument? Actually, could you just give me a workable definition of alignment?

I'm weird and worry about AI x-risk though, so feel free to dismiss me for that if you like.

As someone who is weird, the reason I think these communities exist is because a lot of us are on the spectrum and it's honestly so exhausting trying to discuss ideas with people to just be called weird or toxic or whatever else. In a way the reason people see us as a group is because we can't do the groupthink thing, and so it's nice to talk to others who are not going mock you for liking Harry Potter and being interested in AI risk.

As a "group" we disagree on almost everything... It's why we talk to each other.

> Outputs

> LLMS > Strings / generated text. Strings are flexible and can be anything: chat responses, code, hallucinations, refusals, or even type-safe structured values. To be used by software, responses need to be parsed + validated. There is also always some risk that the AI goes off the rails.

> Jev > Type-safe structured values. Possible outputs and structure are defined in advance. The model never makes type errors. All answers are accompanied with calibrated probabilities and confidence scores.

I mean, this isn't even remotely comparable to LLMs so why compare? Also, why are they bringing up AGI given there approach is so restrictive that what they're building literally cannot have the creativity required for AGI? The video is 100% marketing slop...

The bulk of the application of LLMs is that they generate reasonably reliable text which doesn't need to be defined in advanced. I'm sure there is a niche for this and congrats to the team, but please let's not hype this as if it's the next big thing in AI...

As a Brit I hope Americans understand just how amazing the first amendment is. Here in the UK (and Europe) surveillance can be used to criminalise us for things we say in private. At least in the US this kind of thing is much harder for the government to do.