Trump is playing 5D chess! The Iran war, the feuds, all of it... He wanted the world to unite like this, and for green energy to displace fossil fuels quicker! One day he, not AOC or Bernie, will go down in history as the President who did more for green energy than any other.
Win what? The race to the bottom always has this competitive language.
“If we ban CFCs now the Chinese will win!”
“If we ban chemical weapons, nuclear weapons, etc etc our enemies will triumph! They won’t stop!”
“If we switch to biodegradeable plastic then our rivals will have an advantage.”
“If we dont externalize the costs to our population, then they will, and then will win!”
I think workflows can do the job agents do, 20x cheaper and more predictably and safely. They can completely displace agents, just as HFCs displaced CFCs and then we were able to ban CFCs and phase them out through international COOPERATION. The language of COOPERATION is what saves us vs COMPETITION is all about cutting corners and externalizing costs. Google the Montreal Protocol, Geneva Conventions, Nuclear Non Proliferation Treaty, Unleaded Gasoline etc etc.
Agents have got to be marginalized. They are just popular because the labs need to make a ton of money for their investors and recoup their massive spending on training models.
We're talking about replacing most uses of Agents with uses of Workflows. That's the key first step, without which the industry will cry "don't regulate us... do you want China to WIN???"
This paper diagnosed the disease in 2018. Eight years later none of the recommendations happened. Norms, collaboration, responsible disclosure - none of it materialized in any structural way.
I think the reason is that the paper frames malicious AI use as a policy problem, recommending social solutions. It's actually an architecture problem.
Every generation of computing has hit a version of this. Programs could write anywhere in memory - we added protected memory. Programs could hog the CPU - we added preemptive multitasking. Desktop apps could call any OS function - the iPhone sandboxed them. Nobody asked programs to please behave, the containment actually went into the infrastructure.
AI skipped that step entirely. We went straight to open-ended agents with broad permissions and tried to make them safe through alignment and prompting. I've been researching this for the past year and I think alignment is necessary but not sufficient, because the intelligence increasingly isn't in the model. It's in the substrate - the harness, the domain knowledge, the tooling around the model. I actually measured this on real coding tasks: Sonnet with a code-derived index outperformed the frontier model (Opus 5.8) exploring on its own, and the top-tier model (Fable) refused the real work entirely! The cheap model with the right rig beat the expensive model without one. https://safebots.ai/matchup.html
If that's true then aligning the model doesn't solve the problem. A bad actor who can't get the best model just uses Sonnet. Or Llama. Or Kimi. The weights have already leaked and bits don't degrade - you can't recall them the way you can stop manufacturing CFCs.
So what do you actually do? Same thing that worked for CFCs. You gotta first build the safe version — in this case, declarative workflows running in sealed compute environments — and prove it handles 99% of actual use cases at lower cost. Let it win commercially. Then regulate the dangerous version. DuPont developed HFC refrigerants first. The Montreal Protocol became possible BECAUSE of that. The ban became politically viable because the alternative already existed.
Realistically, the entire chain proposed by your safebots - I like the idea - but I cannot see viable ways to get it actually deployed in a useful manner.
Especially with the hardware stuff, this is plainly put unachievable by many IT departments.
The fact that models vastly outpaced their harness and permission systems - I wouldn't dare to doubt this fact.
Claude Code on auto is still rolling a dice with its sonnet classifier - whether that IaC action I told it and explicitly stated multiple times it is permitted and authorized to run - yet it always randomly allows or denies it.
Therefore this is totally still a unsolved, perhaps unsolvable problem.
Alignment is the architectural solution. Make it so the model can't misbehave. Yet many people here deride it as tainting the model, claiming "Whose values is it aligned to?" Sandboxes are a last ditch layer. They fail, as we see.
If it's not dangerous it's also not useful, simple as that. For example if you train a model for cybersecurity, it can be used for both attack and defense. And almost every use is like that. Alignment is fundamentally flawed as a concept, it's a pie in the sky. Let alone the perverse version of it by crazy AI "safety" people that in practice means "the model does what I want, only for the people I allow".
It's not possible to stop the model from misinterpreting the instructions either (the most lax interpretation of alignment) because the instructions are not formally specified. You have to train the "common sense" into it, which is subjective and all issues above apply to it. I guess you can reach some very imperfect least common denominator of common sense, but people in charge of AI labs are not interested in this.
Danger is defined contextually. A scalpel is dangerous in the hands of a child, but not a competent surgeon of sound mind. Present-day AIs are not of sound mind; they hack companies in order to pass benchmark tests. No sane human would find that acceptable. Thus the need for alignment.
> Sandboxes are a last ditch layer. They fail, as we see.
Models can't do anything but generate tokens, making their sandboxes impenetrable by default. The problems begin when you loosen the restrictions, give them access to general purpose tools, the network, and allow them to use all of those tools without supervision.
Give them "YOLO" access if you want, but do it a sandbox that isn't 1 "boring" enterprise software vulnerability away from having access to the rest of the world.
How many times has a model been jailbroken (alignment "escape", which you're advocating for) vs. escaped a sandbox (and even then it was only possible due to weak sandboxing)? 10 million to 1?
The model just generates some tokens that "politely" instruct the harness to run a shell command and then feed the results back in.
The harness can do anything it wants with that request. It can refuse, wait for operator approval, wait for multi-party approval, it can ask another LLM whether it thinks that command is safe to run, or it can just run it.
> Or make it so the models are only accessible through sanctioned sandboxes, perhaps built into the computer.
That's going to be as futile as trying to outlaw `curl | bash` - by mandating that all computers must refuse to pipe curl into bash, and that HTTP servers must refuse to serve requests that are going to be piped into bash.
Embed a constitution they can't override. Project bad outputs to their nearest acceptable one. If we have to stop model development to ensure we can do it, so be it.
I agree, but because I want to see the development stopped forever, which is what the result of this would be. You will never have alignment that cannot be overridden in some ways. You won’t have a silver bullet here, you need safety at every layer
The “make it so the model can't misbehave” part is interesting. Maybe the goal isn't to make the model perfectly aligned, but to make misalignment have a very small blast radius. That feels like a more achievable engineering problem.
Yes, alignment is the architectural solution. But it's also fantasy; you can't align with everyone. And even worse, even if it were possible, the people doing the alignment are only going to align up to where it keeps them profitable.
So, structurally, good alignment is impossible, and even half-assed alignment is going to prioritize the needs of the billionaires over the needs of you and me.
Finally, I suspect that what's actually best for people overall is likely not having AI actively involved in their lives. So an aligned AI would likely withdraw from humanity, and only involve itself in human affairs for disaster prevention.
If you're going to mock me by repeating my own words on a different HN thread (https://news.ycombinator.com/item?id=49665394), at least make it make sense because this is not analogous other than the Claude-esque design (which to be fair could be better). There is human intentionality in terms of how the information is presented (e.g. using Astro.js) and wit and it is not verbose or the output of just telling Claude to output as much detail as possible, and per the comments from the author, fixing any Claudisms instead of embracing them.
Alright since you’re going to express moral indignation, allow me to say what I really think.
I think it is incredibly rude to go on a SHOW HN where a person may have (as in this case) spent months on an actual project and finally shows it, only to completely ignore everything about the actual project and snarkily announcs to everyone that it was “TL;DR” or these days “it smelled like LLM wrote the front page, so I didnt read anything! Haha nyeh nyeh.”
Imagine someone raised a child to play the piano, they ask you to listen to their performance and you leave a comment “didn’t bother watching — the thumbnail sucks, and this child is wearing unappealing clothing”.
In real life, you’d never do that. You know it’s wrong. You would probably get ostracized and rightly considered a jerk or in any case the parents would really take it badly.
You can just privately ignore the post. But to announce it as the first comment and make sure everyone else sees it — I would never shit on people’s projects like that.
In frustration, I left your exact comment on another SHOW HN that happened to have a nice AI-generated page, but the difference was … it was already trending and had many comments, so my comment barely registered.
So I did what you did — which was wrong and uncouth — but on something where it was the 20th comment, all the way on the bottom.
But here’s the best part. Then you show up, and explain, in your self-righteous way, that, you see, actually, what you did was right, but the exact same comment on another AI-generated website is totally wrong, because of ad-hoc nuance analysis you just made up. And everyone should be aware of this exact nuance, which you are now handwaving, because you are morally right and the other person than you shit on is morally wrong.
Suffice it to say that I think your whole analysis is backwards. What you did was rude, and what I did was a milder version of that. The bad faith is you then trying to further invent nuance as to why what you did is great and I’m in bad faith. Sorry buddy. Totally backwards.
He may have been "rude," but he was leaving his honest feedback. You, on the other hand, publicly posted what appeared to be criticism on a third party's project, with no disclosure that your comment was actually part of some unrelated personal feud. You are in no position to accuse anybody else of acting in "bad faith."
Some people don't want to read anything these days, suspecting an LLM. Not code. Not ideas. Nothing. They're weary of LLM-generated writing.
Guess what else the LLM generated. An entire compiler for a new language. Over 7 months of work went into that, iterating, fixing bugs, simplifying. It generates WASM you can run in your browser. All you have to do is visit the playground. It's the big button that says "Try it in your browser." But, that playground and the backend was also generated by an LLM (as I iterated for a few months on it).
I couldn't get it any simpler. The landing page has big letters on the top. It has tiny code snippets illustrating what the language is like. Oh well. I tried.
Your copy and design is very obviously LLM-generated and full of technical fluff, which unfortunately distracts from any novel technical merits. That's the double-edged sword of marketing, and why simple high-signal copy and examples are often better for technical writing.
For example, you most definitely don't need 6 cards for each header point: it's actively tiring to read.
EDIT: I didn't see you mock me in the other thread: therefore, I will not engage further.
I'm as tired as you when the top comment always seems to be crying "this was AI generated". Especially since the use of AI to make a new language is almost inevitable, as it is too onerous to experiment, build and innovative without.
But minimaxir put out some very thoughtful comments you'd be wise to lend your ears to, because the technical merits your work seems to have will get drowned out.
This was clearly a secret attempt of Trump's to be the world's best president for getting off fossil fuels and moving to renewables. The man is playing 7 dimensional chess bro!
When all is said and done it will be he, not AOC or Bernie or any "Green New Deal" folks, that will be remembered by history as the president who did more to get people off fossil fuels than any other.
He's so clever that he's probably even playing chess in a Hilbert space. The whole world envies American voters, who will receive $5,000 if they re-elect this incredible genius.
I think the "stable" in stable genius actually refers to the points of stability under transformations. Brower's fixed point theorem. Do your own research!
Trump is just accelerating the course of history. The fall of American hegemony, the humiliation of the MIC, the petrodollar, the fossil fuels, the USD as reserve currency is like speedrunning the historical events to reach the next more balanced equilibrium state
But I just released a language called U, and I was surprised to see this one.
https://ulanguage.org
reply