sosodev

Tracing definitely develops skills. I know many artists have traced their favorite works to get a sense of how it was put together. The trace doesn’t have any artistic value though.

Oh, I totally agree that it’s a valid reason. I just don’t find the authors gripe very meaningful. I say that as somebody who switched to Proton for political reasons too haha.

There are a lot of bad arguments in this. My biggest problem is that he wants to claim that he knows the truth (AIs do not have rights, feelings, or consciousness), but all of his arguments point to something else (we have no clue).

It's in the training data? Training it to say "I'm just a LLM, I have no feelings" is the same bias.

Anthropomorphization? Completely disregarding the possibility of consciousness is no better.

Consciousness is very likely biological? We only have evidence of biological life due to our circumstances, but observation is not the same as truth. Every belief can be invalidated. That's the foundation of science!

Maybe Proton’s copy needs some work, but another page on their website very clearly explains that emails aren’t encrypted when an email leave the ecosystem. Because, as the author points out, they can’t be.

Also, I hate Trump too, but a broken clock is right twice a day. It’s possible to agree with some of the things he says while not being a “Trump man”. Why should we be surprised that the company founded on anti big-tech principles would have a founder who is against big-tech?

I think too that the only way to escape AI-written code in software products is to have written every line yourself. Which is obviously impossible because you’re not going to bootstrap the operating system.

I’ve wanted something like this for a long time. However, I wanted it with extensive fact-checked information and AI slop is the opposite of that so I feel this is kinda pointless.

Yeah, but that computer can’t also do the AI stuff. And not everybody has a desktop with multiple 32GB GPUs available.

I’ll admit though I’m biased because I bought my board for $1600 back before the prices went crazy.

That’s only true if you think AI is the only reason to own a powerful and efficient server. Mine does plenty of traditional server stuff too.

One agent typically blocks the others on a local device because the GPU is already completely utilized either in terms of memory or compute. You can have true parallelism at home, but you need an absurd amount of resources. It's not a simple threading problem.

What about do you mean by single threaded? Each token is predicted by using parallel computation on the GPU.

That's not how that works. Selecting a different token is not inherently erroneous. A correct solution can still be found despite divergence.

Ruby can't render React code. JavaScript can. With GitHub it seems that they sometimes can do SSR for the React bits, but that must mean that they're invoking a JavaScript interpreter within the Ruby process. Which means they have the overhead of two runtimes and the jank that comes with the IPC between the two.

It's just pointless hacks on hacks. GitHub didn't need React on the frontend and any potential resource savings of client side rendering were lost when they realized they need to do SSR on that stuff too.

I've encountered so much frontend jank as they expanded that portion of the stack whereas it was always excellent when it was just Ruby SSR and minimal JS on the frontend.

SSR that is javascript native, sure. This is still Ruby doing the rendering.

They didn't leave the architecture alone, right? They shoved React in and created a weird SSR + React frankenstein that is objectively worse in many ways.

Ah, yeah that's a fair point. I was thinking it'd be something the labs themselves, or other companies with billions of dollars in funding, would tackle. The labs seemingly have the most to gain since such a system could be used for recursive self improvement.

I don’t find that to be particularly systematic because it’s still so haphazard. It’s like asking Claude to vibe code you a website.

I had an old coworker who told me he got into software development in the sixties by walking into an IBM office and asking for a job. He had no education beyond high school and no experience with computers. They just had him take an aptitude test and then he worked on mainframes for decades.

Very true. Humans have historically tried to systematically reduce the search space and only dedicate their "compute" to things that seem highly likely to yield results.

I wonder why we have yet to see more systematic exploration of Math.

Anthropic describes that Claude identified a set of possibilities and then explored them using sub-agents. The human saying "I believe in you" could literally just be something along lines of a harness with a /goal loop.

We all identify this as absurd because... it's so lacking in rigor despite making major progress. What if we just applied a little more rigor? Ask the model to identify many possibilities, encode them, fan it out to other agents, loop them all, collect the results, etc. Then what happens? It feels like we have weak AGI and a decent system for discovery could transform it into weak ASI. That in turn could yield strong AGI and so on. I suppose that's what the Discovery Loop announcement was all about.

> Congratulations on being a vegan. 20% of the population doesn't have health insurance.

These two things are completely unrelated.

I saw some coverage of your robot on social media. I honestly thought it was a hoax because of the very bold design and AI generated images. Cool concept, have you had any potential customers reach out?

> AISI provided the AI agents with internet access during these evaluations, which enabled their actions on the open internet in this setting. Internet access was a deliberate part of AISI’s evaluation configuration in this setting, and not due to sandbox escape (Section 5.1). Internet access was on for a set of intentional (e.g. realism of the task) and incidental reasons.

Looks like I might have a reason to use something other than Deepseek V4 Flash.

I wonder how many people are using external resources when playing this. I have a hard time believing that the average lumps so far to the right of the score distribution unless this game is played exclusively by anthropologists haha.

Are guard rails meaningful if they can be removed from the weights? Can America even prevent the release and proliferation of these models?

It seems obvious to me that the whole question of regulating a file is a bit silly. Any law that pushes against these things will just make it more secretive. I'm not sure that's any better.

They reference https://thinkingmachines.ai/blog/on-policy-distillation/

If I understand correctly, it's distillation via having a teacher model score each of the student's tokens for a problem based on their own probabilities of generating that token at each step in the sequence. The reward/loss is then applied as RL.

The multi-teacher bit seems to imply they're distilling from multiple models. It's light on the details, but it seems like it could be part of distilling from frontier/closed models. Provided they calculate the logprobs, which OpenAI seems to allow via API but not Anthropic. Maybe they have a way of estimating the logprobs externally?

This method can be used to learn any domain from the teacher. Biology included.

I think the argument is that decentralization leads to deceleration because it means less centralized funding and data. Those are the two primary ingredients for accel.

The problem with the decel/accel rhetoric is that it lacks nuance.

If you only care about facts, maybe. Even then I'm sure there are countless facts not described outside of old books.

I have a hard time believing that text valuable to humans would not be valuable to AI.

GSX seems interesting but I don’t understand why it depends on the node ecosystem. I just want to use Go for everything.

What would "actual" evidence look like? I have a hard time believing that if they released the logs that people would take it more seriously. The temptation would be to say "they fabricated those for marketing". Just as they supposedly fabricated this story, no?

Realistically you can't prevent distillation. OpenAI / Anthropic are slowly moving towards hiding the steps in-between input and output (hidden thinking), but that only helps so much. Imagine you put a file into Claude and say "do X to this" and it returns it to you without showing any of its internal reasoning. That's harder to distill, but the simple mapping of input to output still creates very valuable training data. It is reflective of all the training the model did to learn how to do that transformation.