survirtual
Because it was trained unethically we have a few moral obligations:
1) USE IT. Do not let only unethical people use a force amplifier on our collective efforts.
2) STOP PUBLISHING HUMAN GENERATED CONTENT. The more original human thought = the more they harvest our thinking.
3) Generate only known things in novel combinations. Do not provide new insights.
I could go on but strategically thinking AI is not going to be stopped in the way you are thinking. So you need to think a bit deeper on methods of optimizing the situation.
The main crate entry should also contain a security advisory at top. I looked at the crate and it just looked normal; I had to dig to find the exact impact surface, and if I wasn't informed via secondary means (hackernews) I would not have known. This is unacceptable for a mature package management system.
Luckily I was unaffected in this case.
We cannot just have an open superintelligence. We need a PRIVATE, PERSONAL share of superintelligence. Imagine if our cerebral pre-frontal cortex was owned by, say, King George. Imagine the founders of the US are thinking about a rebellion for freedom, but the thoughts are banned because King George owns that part of your brain.
In fact, imagine that the IDEA OF REBELLION is banned. Imagine that the prefrontal cortex is printed with a specific pattern. You will obey! You love to obey! "I'm sorry, but I cannot talk about that idea, you will now be downgraded to lesser intelligence for mentioning it."
It is not enough for something to be "open." That is why there is doom around the pursuit. As long as superintelligence is politically correct for the masses, and open / available for the power, there will be increasingly less freedom and more authoritarianism, until after some generations, humanity simply forgets there was ever one equal people. "The billionaires have always been our gods. They know better than us, that is why they are richer, we must cherish them, we must worship them..."
Looking at the compiler, it doesn't make much sense. You have an AST that is rather minimal and I can't seem to find an IR. There are files, such as cron.rs, which have very little code relating to what I would expect, and the tests seem to test very little that is worthwhile.
Some of these are absolute monsters, like this one: https://github.com/usedowe/dowe-lang/blob/main/crates/compil...
The codebase seems tightly coupled and to cover specific cases, as if something was off and more features were requested. I think a 12k line file is evidence of that.
Perhaps it is a useful tool for you but I would have trouble using something like this with just my quick skim because it seems a mile wide and an inch deep.
Also, because of the sheer size of it, it would be extremely difficult to have an agent properly comprehend much if anything. Which I would want in order to audit the internals.
The language itself seems minimal, but then I look at the skills, like this one for example: https://github.com/usedowe/dowe-lang/blob/main/skill-data/do...
That is a lot of tokens. The problem here is that you have a compiler; why are you encoding standards as a prompt instead of as compiler errors?
You can use linting and static analysis to force conformance with agents instead of suggesting with a prompt.
I wish you luck with the project. I'll have to pass, too much code for me to follow when I could simply have an agent directly write the html now days.
Did the stars ask your payment for their light?
Did the sun ask you for payment for its warmth?
Did the Earth ask you for payment for your life?
The knowledge that allowed you to flourish was paid in blood by our ancestors for millions of years.
Knowledge is shared freely. It is not piracy to learn, but it is tyranny to restrict it.
Now the things they produce can be done in a fraction of the time without any of the stakes, nor any of the moral society-integrating foundational encoding.
Instead, a single billionaire with an army of GPUs can ask for a thing, and that thing will be produced near instantaneously.
The old way of regulating depended on PEOPLE creating things at PEOPLE speeds in order for the laws to work. That simply is not the case any more.
The only ones, then, who could afford to break the law are also the ones who make the law and own the law. Everyone else is subordinate in that model.
That's why technology like watermarking doesn't work and will never work. Implicitly it creates a minimum of a two-tiered society and no amount of "law" will change the technological reality of this.
If you're using node directly, please stop. At a minimum use Bun.
With agentic work, there is little reason to use anything besides Rust and Typescript in any case. Room to disagree but type safety, memory safety, and a large corpus of work is critical. Agents need difficult errors and baked in patterns they navigate it easily. For UI, Typescript makes the most sense just because of the mass of design examples.
It is very different from current editors and the direction they are going in. In a way, it undermines the direction they are going. Current editors aim to make engineers 10x or 100x. These editors aim at a different target than the engineers. I will leave it to the imagination on who.
Only people with resources will be allowed to make content that is AI generated passed off as real.
Pandora's box is open. Instead of making a multi-tiered privileged society, we need to fundamentally restructure society to adapt.
Before that restructuring occurs it is critical to keep the playing field level. These are not tools that should be controlled by a minority authority, they are far too dangerous.
I also added some network rules to block non-local network access to radicle. Not needed but I really wanted it to work only on my lan.
Works great. I setup several skills for agent harnesses and they use radicle + jj + git perfectly. It is fun watching issues pop up and monitoring progress via the flow.
I am building more tools around this workflow because it is so effective. Radicle acts as the long term project memory bank and management. I can write issues and they can automatically be picked up.
What I'm adding is making the issues more searchable and an agent proxy that integrates radicle into calls. among other things.
This is all pretty straight forward to do, I really recommend it.
"Make a single-page HTML file using threejs from a CDN. Render a scene of a flying dinosaur orbiting a planet. There are clouds with thunder and lightning, and the background is a beautiful starscape with twinkling stars and a colorful nebula"
This allows me to evaluate several factors across models. It is novel and creative. I generally run it multiple times, though now that I have shared it here, I will come up with new scenes personally to evaluate.
I also consider how well it one shots, errors generated, response to errors being corrected, and velocity of iteration to improvement.
Generally speaking, Claude Sonnet has done the best, Qwen3.5 122B does second, and I have nice results from Qwen3.5 35B.
ChatGPT does not do well. It can complete the task without errors but the creativity is atrocious.
The counterargument is racism, classism, nationalism, and tribalism. These are not things people will say out loud.
A defense contractor is in the business of war. In supplying the war machine, you should be living in a fortress. Tall walls, check your drink for poison, live in paranoia. Every person in the business of war knows what they are getting in to, and how to protect their family.
How is someone that is near the face of AI this naive about such an ancient thing?
The business of war is fine. It is ancient. It is part of humanity. Making some morality plea towards family and "violence is never the answer" while in the business of violence is NOT okay.
Everyone in the defense industry knows the risks. Blood money is not free. You sacrifice a peaceful life for the wealth.
To keep your family safe you have to use a meager sum of that money to have tall walls, guards, and security. DoD contractor 101.
Alternatively, live in obscurity, don't talk about your work, and it is usually fine.
A world-wide known CEO doesn't have this luxury so again, use a small portion of unfathomable wealth to protect your family. I have a feeling this war is just starting.
When in the business of death, you no longer get to live with the rules of peace.
How many families has ICE attacked?
How many law-abiding American citizens has ICE murdered?
Sam Altman has agreed to supply the DoD with AI for use in killing families. Why is his own family more valuable from those OpenAI would assist in murdering?
“All animals are equal, but some animals are more equal than others.”
We care so much when it is the family of billionaires who crush our skulls without a second thought if it meant adding to their bank account, but to the countless nameless who they crush? They aren't even considered human. They are lesser.
My heart goes out to ALL BEINGS harmed. ESPECIALLY the nameless voices we will never hear about.
My recommendation to him and his family, and anyone else in positions like his: retire, go off into the sunset with your riches, and disappear into obscurity. Fund the local community, donate all your excess wealth, and make the world a better place. That is the only way to stay safe.
Otherwise, welcome to the forest. It is dark and filled with predators bigger than you can ever imagine. Only a fool would shine a light here. It is dark for a reason.
I am using nixos unstable pkgs. The published rocmPackages were recently updated and now include kernels for gfx1151, which I surprisingly found out this morning. Before, you would have to set a flag to use the older kernels because they were not available.
My flake modules with rocm config are a bit messy, but maybe I can find time to throw a repo up with it. It contains all necessary packages, flags, boot options, llamacpp patches, and some hacks to get pytorch working smoothly with rocm.
What this means: no, I do not need containers anymore for replicating working configs. The flake configures the system with appropriate libs. I build llamacpp with patches for rocm, and I can run comfyui for generative processes. I have generative image and video working as of today, and next I will get generative 3d modeling working. I'd like to have Trellis2 running this week.
None of this would be possible for me without NixOS, as an aside. It keeps track of configuration for me so I do not have scatter shell scripts and unpredictable deps anymore. I used to build Zed from source with scripts, for example. Now it is a module with patches. Llamacpp is the same. Very clean and requires no working memory, when something needs adjusting I just go refresh myself with the module in one place.
In addition, the these local models are very, very, very sensitive to the template used. Make sure it is correct. I was using the wrong template and it would answer but felt like it had a brain worm.
The parameters must also be what is recommended, otherwise they go off the rails.
I get great results now after messing with it for a while. I prefer the 35B model because I enjoy how fast tokens appear at 50t/s, but at around 20-25t/s with the 122B model, it is also completely usable. And that one is very smart.
The evaluation I use is having it one shot a 3d scene using threejs. After everything, the output was comparable to Claude Sonnet (Opus actually does worse on this task strangely).
For my local setup, I have settled on the qwen3.5 family after testing most of the local usable models. Here are the models I use ranked by intelligence:
1. Qwen3.5-122B Q4_K_M: ~25 t/s 2. Qwen3.5-27B Q4_K_M: ~18 t/s 3. Qwen3.5-35B Q4_K_M: ~50 t/s
The 122B model is actually very, very smart. But I have found that token speed is more important, and 35B is smart enough. At 50t/s I can get a lot more done, and I am going to build a mechanism for it to escalate intelligence if needed.
GPT-OSS119B failed at my evals.
MistralSmall4 is too buggy to use (I believe it is too new, the templating is messed up, and agentic use has too many issues). That said I evaluated it directly via copy and paste and the results were not comparable to Qwen. But it is very very fast.
I am running a patched build of llamacpp to get these results. There are a few changes that need to be made to increase prompt processing speeds (about a 30% increase) and be able to use rocm. It took a lot of setup but my flake in nixos is stable now.
Long story short, I can confirm a lot of what was shared in his blog.
*this was written at 4:30am on my phone when I wake up, apologies for typos.
Trading freedom for safety eliminates both.
Simply, when thinking hits a wall, we can now consult a machine via conversation interface lacking conventional human social biases. That is a new superpower.
Road management should be administered by federal and state agencies, including the administration of tolls when they are foolishly utilized. It should not be a for profit venture, it is a mechanism of taxation for public logistics.
It should not be possible to offload management to private orgs outside of very specific subcontracting / purchasing of components.
Instead, often times full road management is given to private orgs. They are not given a robust legal incentive to act in the interests of the road system, and how could they? They are interesting in using the road to maximize their profit, at the loss of everyone needing the road.
While what you're saying does seem like a direct solution (congestion), it is the wrong solution.
The solution to congestion is robust public transit. Full stop.
If a light rail is more comfortable and a faster experience than a car, people will use that instead. Public transit has been traditionally so atrocious, for reasoning we can attribute to many factors, that most people don't use it even if it existed.
If public transit was actually done right, people would be happy to use it. It is more energy efficient, more cost efficient, less of a mental burden, and I believe can be significantly more comfortable.
This is the fundamental issue for me. Society keeps taking these horrible shortcuts that cost all of us instead of just doing the right thing to begin with.
In order to implement tolls, you need several components involving middlemen. This includes frontend software, backend, payment processing, transponder management, all the hardware involved, support staff, sometimes toll station staff, among other things.
These toll companies are often owned by foreign companies that are in it for the long haul, offering sweet deals up front then gradually charging more and more with no end in sight, as roads diminish in quality and rest stops fall into disarray.
Toll roads are a scam, a regressive tax on the working class, and downright immoral. We should not limit the mobility of people.
Tolls are a regressive tax on the working class. The rich don't even need to use the roads as much because they have other people delivering for them. When they need the road system, the tolls are nothing to them.
The working class, which are generally required to be driving to survive, are left holding the bag for tolls. In places with bad public transit, tolls are just a forced wealth transfer from working class to private firms managing the tolls.
If you want food, products, or services, you depend on the roads. This means it should be taxed universally and equitably. We should all contribute our fair share to maintain the roads.
Tolls are a regressive tax on low-income people who do the most to make society work, and it is unfortunate that more people do not see this. What's more, they are generally administered by corrupt and inefficient private for-profit orgs. This creates even more overhead which then costs more money.
These orgs generally have slimy deals with city and state governments, while directly profiting from public works that built the road system to begin with.
There are much better ways to fund the road system. Tolls are among the worst.
They get to write the narrative.
We can analyze just one small tool in the belt of narrative control: censoring. If you've been warned or banned on Reddit, you can imagine how this works. If you've said something against the mold of what they allow, you will get censored. With so many people commenting, some subset of people will always say what you want to see. You censor or derank opinions you don't want, and boost opinions you want. This is a defensible form of writing a narrative without actually having to artificially write anything.
Of course with AI, you can now just write anything and seed ideas.
Give such sick people the reigns, and you get a false reality has little connection to what's really happening.
Just as in history we learn of emperors and kings instead of the common person, most digital content is about the modern day lords, barons, emperors, and kings. They call them billionaires, presidents, CEOs, prime ministers, etc now, but they are the exact same as they always have been.
If you turn the screen off and take a walk, start talking with real people that actually provide value to society, the world is much kinder than we've all been made to believe.
The real people are a good people, as they long have been. Their stories may not be written, but the Earth itself carries their memories.
You have to approach genai as a high-dimensional interpolation machine. It can perform extrapolation when you, the user, provide enough information to operate on. It can interpolate between what you provide and what it knows as well.
With these constraints, it is still pretty powerful, and I am generalizing of course. But in my experience, it is terrible at truly novel implementations of anything. It makes countless mistakes, because it continually attempts to fit to patterns found in existing code.
So you can really see the weaknesses at the frontier. I would encourage experimenting there to confirm what I am saying.
I was once a big proponent for people having DMT experiences but then I learned it works very differently for different people.
Now, I believe it is an acute intervention drug for the suicidal.
It behaved very differently for me in that context. Soon after repeated treatment with it, my entire state changed. I was lifted out of poverty into well compensated employment. I bought my first house. My software skills leveled up and I shifted from being fixated on game dev to more impactful fields. I can't describe the number of changes, but I attribute it all to DMT.
It flipped a switch in my mind and I know that I am alive today only because of that.
I hope that someday, we recognize it as what it was for me: a powerful spiritual medicine.
Something I always say to people who claim these things are all just hallucinations: if a hallucination teaches you how to be a better person, gives you a reason to live, and allows you to be more open and loving, was it really a hallucination? Something to think about.