The memory problem is already being addressed in various ways - antigravity seems to keep a series of status/progress files describing what's been done, what needs doing, etc. A bit clunky, but it seems to work - I can open it up on a repo that I was working in a few days back and it seems to pick up this context such that I don't have to completely bring it up to speed every time like I used to have to do. I've heard that claude code has similar mechanisms.
I've been doing stuff with recent models (gemini 3, claude 4.5/6, even smaller, open models like GLM5 and Qwen3-coder-next) that was just unthinkable a few months back. Compiler stuff, including implementing optimizations, generating code to target a new, custom processor, etc. I can ask for a significant new optimization feature in our compiler before going to lunch and come back to find it implemented and tested. This is a compiler that targets a custom processor so there is also verilog code involved. We're having the AI make improvements on both the hardware and software sides - this is deep-in-the-weeds complex stuff and AI is starting to handle it with ease. There are getting to be fewer and fewer things in the ticket tracker that AI can't implement.
A few months ago I would've completely agreed with you, but the game is changing very rapidly now.
this works fine for like 2-3 small instruction sets. once you start getting to scale of a real enterprise system, the AI falls down and can't handle that amount of context. It will start ignoring critical pieces or not remember them. And without constant review AI will start priotizing things that are not your business priority.
I don't agree they have solved this problem, at all, or really in any way that's actually usable.
What I'm saying is, don't get to thinking that the memory problem is some kind of insurmountable, permanent barrier that's going to keep us safe. It's already being addressed, maybe crudely at first, but the situation is already much better than it was - I no longer have to bring the model up to speed completely every time I start a new session. Part of this is much larger context windows (1M tokens now). New architectures are also being proposed to deal with the issue, as well.
Context windows are a natural improvement, but new architectures are completely speculative and it’s unclear we can make any sort of predictable progress with new, better architectures. Most progress has been made on essentially the same architecture paradigms, although we did move from dense models to MoE at some point.
The examples you give for jobs of the future don't sound appealing or very numerous. It seems like you're saying that people will be employed as personal assistants to the uber wealthy. But there aren't a lot of uber wealthy - certainly not enough to employ large amounts of the economy.
I mean, there's always the job of building pyramids. But no, seriously, I don't think it's just about the ultra-wealthy. Basically, anyone better off than you. Which is basically what's going on today: you effectively work for your boss, they work for their boss, and their boss (possibly after some extra hops) works for the ultra-rich CEO.
Datacenters were very automated when RAM was infinite. As the world becomes compute-constrained, the economics may increase the demand for smart hands mixing-and-matching server components to turn two broken servers into one working server.
Was starting to think about setting up a neighborhood Signal group, but now thinking that maybe something like Briar might be safer... only problem is that Briar only works on Android which is going to exclude a lot of iPhone users.
Why wouldn't you just use random abandoned forums or web article message threads? Iirc this is what teenagers used to do when schools banned various social media but not devices. Just put the URL in a discrete qr code that only a person in the neighborhood could see.
I spent a dozen years in SF, where my friend circles routinely used Signal. It's my primary messaging app, including to family and childhood friends.
I live in NY now. Just today, I got a message from a close friend who also did SF->NY "I'm deleting Signal to get more space on my phone, because nobody here uses it. Find me on WhatsApp or SMS."
To a naïve audience, Signal can have a stigma "I don't do anything illegal, so why should I bother maintaining yet-another messenger whose core competency is private messaging?" Signal is reasonably mainstream, and there are still a lot of people who won't use it.
I suspect you'll have an uphill battle using something even more obscure.
> The more liberal, the older they are. With conservative/liberal being adherence to scripture... The more liberal, the older they are. With conservative/liberal being adherence to scripture.
And yet many young Evangelicals have deconstructed and dropped out of those conservative congregations over the last 20 years or so. They couldn't bridge the cognitive gap between the conservative political stance of their church and what they read in the bible.
It's not always obvious anymore what music is AI generated. And it will likely get harder to determine. I wonder if this kind of policy is going to lead to artists who are not using AI to be accused of doing so?
> we will be the harbinger of a large-scale societal collapse.
The harbingers are already here and have been for a while now. Harbingers are signs that things are breaking down. More like "or else we will have a large scale societal collapse".
I've been doing stuff with recent models (gemini 3, claude 4.5/6, even smaller, open models like GLM5 and Qwen3-coder-next) that was just unthinkable a few months back. Compiler stuff, including implementing optimizations, generating code to target a new, custom processor, etc. I can ask for a significant new optimization feature in our compiler before going to lunch and come back to find it implemented and tested. This is a compiler that targets a custom processor so there is also verilog code involved. We're having the AI make improvements on both the hardware and software sides - this is deep-in-the-weeds complex stuff and AI is starting to handle it with ease. There are getting to be fewer and fewer things in the ticket tracker that AI can't implement.
A few months ago I would've completely agreed with you, but the game is changing very rapidly now.
reply