civvv

Jesus Christ, so much "stuff" crammed into so little space. What a nightmare of a website.

Is it faster? Thats the only thing that matters for me anymore. Models are probably as good as they ever need to be for coding, but please make them faster, please.

> We could get what we need to survive on a fourth of the farmland if we kicked the addiction to meat

100%

> arguably a more malicious crime against animals and the planet

Personally disagree, it entirely depends on your moral and ethical alignment though, of course. Generative AI will be used to divide and conquer us at a pace never seen before, it will destroy the internet/digital realm by burying it in generated content.

Doesn't matter to Bob, who lives across the road from the datacenter and has an artificial sun outside his bedroom 24/7 and wants to burn the thing to the ground.

The only relevant difference between datacenters and agriculture is that we need one to survive as a species, the other is actively being used for malicious crimes against humanity.

This is true for code as well. Code is the specification, writing it in English and having a machine predict the details will only give you a worse result.

Welcome to the new internet. It was fun whilst it lasted. Next evolution will likely be closed, invite only forums.

This comment section is very astroturfed.

Same probably goes for slot machines if you strip them of all the bright, flashy colours and comforting sounds and make it a black metal box, with only a number display.

Its almost like... it was inherently designed to be as addictive as possible?

Genuinely curious, what is the benefit of doing this over just programming it yourself? It sounds to me like this is just an extra step that will grant you less control?

When programming a feature, you are simultaneously doing at least four things: 1. Implementing. 2. Building highly detailed mental models. 3. Learning and expanding your skillset. 4. Quality control and scope limiting.

And this process can be iterative and dynamic. Writing massive, super detailed specs that you then hand off to a undeterministic model feels like doing step one and three, while skipping two and four, which you then have to do after. What is the benefit? The speed up, in my opinion, comes if you skip step two and four, but then your product WILL be worse. Feels like I am going crazy?

Programming was never the bottleneck for software dev?

Yeah, thats likely true to some extent. But my point is perhaps a "counter-thought" to the idea that this is ASI level intelligence and that math is now a computer task and not a human one. If it is incapable of moving maths forwards into the "unknown" it simply can not replace humans, and this might be a hard structural block for LLM's.

Except it is very easy to verifiy whether or not someone has used a nuclear bomb, where as LLM's can be used in complete secrecy. MAD only works if you can verify that the other part is not using it.

How have we got to the place we are today then? Someone must have made the first steps onto uncharted territory, otherwise we would be in a homogeneous state frozen in time.

I am not saying that LLM intelligence can not be the same, that they are uncapable of dicovering new fields/problems that they have no training on. I am just pointing out that historically it kind of "must" be true that humans are capalbe of this, but we have yet to see an LLM do something like this, something radically "new" in a sense. All of these breakthroughs appear to me (not a mathematician) to be more a case of "digging" through millions of existing attempts/work, patching it together into a result.

This would already make LLM's one of the greatest tool mankind has ever made, but it has yet to display what I would consider a necessity for human level intelligence, which is this ability to discover entirely "new" things.

Would an LLM, given enough time and only the currently available trainingdata with no further input from humans, be able to solve something that was discovered tomorrow?

For humans my answer would be: maybe, probably, because this has been done historically.

For LLM's I would not be comfortable in claiming that they could. I think they would not be any better at this than traditional computational bruteforce.

LLM’s seem very good at solving mathematical problems of which there is an enormous amount of exisiting work/attempts in their training data. This is an amazing capability, but does not convince me that these models are «thinking» or «reasoning» in the way a human does. A human mathematician could in theory categorize/discover an entirely new field of mathematics tomorrow, based purely on their «human intelligence», I wonder if we will see similar examples by LLM’s soon. It seems to me currently impossible that LLM’s can replace human mathematicians, because of their (assumption) likely dependence on human input in the sense of enormous amounts of pre-existing attempts/data.

If an entirely new problem, within a new field of mathematics were to appear tomorrow, I highly doubt an LLM would be useful at all on their own. Is this the «ultimate ASI test»?

I’m running Qwen3.8-27B-Unleashed UD-Q3_K_XL, which is a ~12.3 GiB Q3 quant, fully offloaded to the 16 GB 9070 XT. I disabled the vision projector to save VRAM and use one inference slot, Flash Attention, Q4 KV cache, --fit off, and --ctx-checkpoints 0. I’m running it with a 64K context window. The AMD driver also needs to be recent enough for ROCm 7.14; I targeted Adrenalin 26.6.4 or newer.

I got Qwen3.8-27B running on my RX 9070 XT by setting up AMD’s ROCm 7.14/TheRock runtime for gfx1201 and adding the required HIP/LLVM paths to Windows. The official llama.cpp ROCm build detected the card correctly, but kept crashing with 0xC0000005 while loading the model, so I switched to the Windows Vulkan build of llama.cpp instead, which worked reliably.

I’m running Qwen3.8-27B-Unleashed UD-Q3_K_XL, which is a ~12.3 GiB Q3 quant, fully offloaded to the 16 GB 9070 XT. I disabled the vision projector to save VRAM and use one inference slot, Flash Attention, Q4 KV cache, --fit off, and --ctx-checkpoints 0. I’m running it with a 64K context window. The AMD driver also needs to be recent enough for ROCm 7.14; I targeted Adrenalin 26.6.4 or newer.

Running Q3 on my AMD RX 9070XT. 32k context and 32/TPS. Apart from the context window preventing it from doing any large tasks, this thing is seriously powerful. I could probably push it to 64k context. Local open models are the future, and I am definitely getting a more powerful card. Very fun!

Hardly an accurate statement. Harnessed LLM’s are clearly capable at finding an answer to a mathematical problem in a pool of existing attempts by mathematicians, but I highly doubt that it will be able to solve math that it does not have massive amounts of pre-existing training data for. Which means we are entirely dependent upon intelligent mathematicians to continue their work and implement LLM’s as what it is: a great tool.

If I was a supervillain working towards the extinction of humanity I probably wouldn’t boast too loudly about it. Might anger, lets see… at least a couple billion people?

Guess its back to the good ol' map, compass and stopwatch.

This has been done by the open-source community for a decade+ already. https://github.com/LoopKit/Loop

My younger brother has been using the open-source system Loop for his diabetes T1 since he was six years old. It automates the bloodsugar monitoring and insulin delivery. All running on his iPhone. This technology has permitted him to live almost as if he did not have the condition. Truly revolutionary, and the project is being developed by an amazing community of volunteers (total heroes!).

When does the intense planning just become programming in a natural language though? And at that point, it just becomes less accurate coding. I feel like this is the biggest trade off when it comes to agentic coding. You are practically trading accuracy and control for speed and efficiency. And if you want the same accuracy and control when using agents for coding, you as a validator become the bottleneck, removing in most instances all of the speed and efficiency gains. It all comes down to this, what matters the most: speed, or accuracy and control. Todays capitalism typically favours short-term thinking, speed and efficiency, which is bad news for everyone. Software developers will be forced to have less control and produce less accurate code, because the validation step is not permitted, due to it being the bottleneck.

Yeah, I host three small-scale game servers for an indie-game, all out of pocket. 50-80 players hop on every day and the community is super nice. Its been possible because the price has been €16 per instance, but with these new prices, it will absolutely not be possible anymore. I imagine a lot of small game server communities will meet the same fate now, since Hetzner has been the go-to for this kind of stuff.

>Or as Jean Baudrillard has said:

It is nothing short of profoundly ironic to quote Jean Baudrillard in this context.

Yarvin is a basement dweller and 4chan intellectual, high on his own supply of pseudo-intellectual takes. What is sad and worrying is that these kinds of politics are increasingly moving out of the fringe internet and into pockets of power (eg. Thiel and Vance). It is problematic that these ideas now linger only one or two steps away from the most powerful and influential person in the world.

There are many indications that model progress is slowing down, so that is not entirely accurate.

Three of my favourite game series as a kid, what a legend.

Not sure about their other departments, but at least in March there was no LLM coding in their Source2 & game teams.

https://x.com/ZPostFacto/status/2035784300575305895

I am no insider and have never even tried to build an LLM, so I can only guess. But the general sentiment seems to be that this is the case. If you are interested, I would recommend you read the MIT paper "Superposition Yields Robust Neural Scaling" [0]. It confirms an interesting trend: models represent more features/concepts than they have clean independent dimensions, so features overlap. Increasing model dimension reduces this geometric interference, which lowers loss in a predictable way, but with diminishing returns.

This has, in my opinion, likely been the primary vector in getting better models thus far, but MIT mathematically proves that it yields diminishing returns for each new dimension added. It will get more and more expensive and the cost-return will or probably already has made it infeasible.

Ilya appear to support sentiment this as well. [1]

[0] - https://openreview.net/forum?id=knPz7gtjPW [1] - https://www.businessinsider.com/openai-cofounder-ilya-sutske...