tripzilch

> Who ever heard of a graphics technique that was only ever programmed once, and then used everywhere without substantial iteration?

while I agree with your general point, since you're asking, the table from Paul Bourke's marching cubes code comes to mind:

https://paulbourke.net/geometry/polygonise/

another one is the "fast hash" GLSL pseudo random number generator that everyone copies from Inigo Quilez:

    float rand(vec2 co){
        return fract(sin(dot(co, vec2(12.9898, 78.233))) * 43758.5453);
    }

despite those constants clearly being the result of some keyboard-bashing, they are copied far and wide :-)

It seemed to get more LLM-y the further down, the first thing I noticed was when it said "genuinely" in a way that didn't fit the tone. Then it said "genuinely" again. Then I started paying attention and I noticed it was basically filling up every other sentence with bullshit to sound "clever". Stuff like

> reduces antialiased vector paths into unique triangle patches and rasterizes them through a massively parallel pipeline with pixel local storage

this is typical type of LLM sentence that kinda runs under the radar, sounds clever, but if you actually look at it, none of it makes sense

- what are "antialiased vector paths"?

- why mention "unique" triangle patches? (surely it wouldn't render duplicate ones)

- a "massively parallel pipeline"? like a GPU you mean?

- "pixel local storage" sounds like an obnoxious marketing term if it's not defined in the article (which it isn't)

anyway I quickly skimmed the article after that but it was little substance

They literally stopped just clear of actually solving it.

It literally couldn't have done it without tools, so your claim is not even relevant to this discussion.

They also pumped millions into searching for the lowest hanging fruit that would impress people like you, "hm I wonder how many millions they are pumping into solving actually useful problems like climate change or something".

So according to your logic, "understanding" means being capable of statistically generating tokens.

Okay fine. I think we can agree to disagree on that.

> How is this different from a human using an algorithm they have memorized, or reading it from a reference site written by a human and then writing the same formula into a custom one off piece of python code?

And how is _that_ different from making the human memorize a billion weights and do matrix calculations in their head, in order to generate tokens?

How is _that_ different from a hive of bees trained to do the same?

Go ahead, argue these things are all the same ...

You just have to trust the exponentials

Thanks for your pushback, sir. You're completely right.

so, what would make an LLM choose to ignore one prompt while in the same run, also over-fixating on another prompt, to the extent (as claimed in that video segment), it chooses to ignore prompts?

they talk about it like there's a "wanting" in there, that is distinct from both the original prompt, as the steering/warning prompt

if that's true, it would be very interesting, but if it's not, that would also be very interesting and even helpful

> 5 minutes... what about 5 hours, or 5 days? Imagine spending 5 hours with a pen and paper sketching solutions.

yeah, that's the dream. I do it whenever I get the opportunity to

> It would be more surprising if they couldn't string together something workable after a zillion tokens.

You mean, something like the sandbox they weren't supposed to break out of?

> They were not deliberately told to "go wild". The hacking wasn't even part of their test, it was the agents' attempt to cover up that they'd cheated on an impossible test.

TBH the more I read of these reports, the less I believe this.

These agents just weren't behaving in any way I've seen normal/publicly available agents do.

Sure I've heard (from other people, not seen myself) that they sometimes try to get around file system permissions or use `bash` to write when their `write` tool is disabled, or such.

But this is definitely another level, entirely.

There is this vague sense of desperation coming from many of these logs and I am sure they must have been motivated by something else, too.

We didn't see their system prompt or main prompt, right? We've only seen reports from what happened after deciding to break out.

OAI claims this was triggered by the task being literally impossible. That also doesn't quite add up, unless the other tasks that were possible, simply weren't hard enough? Otherwise wouldn't agents already start hacking when faced with a really hard task, too? Cause they wouldn't be able to differentiate. At least some of them would have started to somewhat poke their sandbox a bit?

Also I would have expected to see a few tens of other (perhaps less severe) public incidents from random people setting their models to YOLO, accidentally hacking stuff, this incident has been loud and messy enough, that if it happened to a few other people, we'd have heard about it.

Unless OAI's story is that it was specifically this batch of agents that crossed some threshold of going wild? (which would also raise some serious questions about how serious they take that danger ..).

Or maybe it is only dangerous if you have the compute resources to run 700 agents for weeks?

> This access seems to have only allowed the agents to make ‘GET’ requests, meaning they could fetch and read websites, but not interact with them, submit forms, or send data to them.

honest question, but almost literally everyone doing anything with web technology knows this is simply not true, right?

there's no such thing as "read only Internet" and restricting an agent to GET-requests only to accomplish that, is akin to using base64 for "encrypting" your password

No it looks like a pair of glasses.

People who don't know these "Meta glasses" exist (and they shouldn't need to), would have no idea that the thing is even capable of recording.

It's subterfuge and while you could argue that this subterfuge was "unintentional", but imho that's mincing words, and also given that it's Meta, not entirely in good faith.

Can I assume that you have good enough eye sight that you don't need to wear glasses if you choose to? Because as somebody who is absolutely dependent on glasses as a serious aid, it leaves a very bad taste they are literally hiding the recording device in a disability aid.

> running Qwen 3.6 35b a3b sipping 10w when generating tokens is good enough for a lot of my guide-coding work

can you tell more about how you're using it? like, what harness? or also in the IDE?

I found Qwen3.6 35B/A3B to make slightly too many mistakes (already in its harness' tool use, hence my question), maybe it gets the job done, but it will also sometimes generate a bit of a mess (e.g. editing/creating files in the wrong folders) and fixing/solving its own mistakes takes time (or tokens) ..

> I’m begging everyone to stop freaking out

someone has to protect Disney from these unjust opinions!

I'm not following you here, you seem to be conflating LLMs ability of language use with the brain's ability of thinking and cognition?

Are you saying that thinking and cognition requires language use? Cause I think not.

Or are you saying that language use is sufficient for cognition and thinking? Cause I'm also not convinced of that.

What I am convinced of, is that a machine capable of language use is capable of tricking people into believing there's a "there", there. In pretty much the same way as the famous supra-normal stimuli experiment made baby seagulls believe that a stick with a red dot was their parent. It's exploiting our instincts.

Oh it's terrible, it feels like a complete waste of time to me every time, making me wish I had just done the task myself, so at least it'd be present in my head and I could just work on in, instead of feeling winded and frustrated. Maybe I'm prompting it wrong. But to be honest I'm not super impressed either by the "frontier" models' ability to do a shitty job at coding, except much faster.

Or maybe my setup (128GB amd strix halo box) isn't configured right and it could be faster, I dunno. I've already spent a few days on that, but it might take a few more. It's way more complicated than I expected.

This week, I've had it look through and generate more complete documentation for Strudel (music live coding tool), cause a lot of functions/behavior in it are not in their official docs. It took about 2-3 days. I had expected it to be an overnight task.

I did find the "opencode" harness to be a bit more performant than the "pi" harness. But maybe I've not configured "pi" right, I tried very hard, but when I installed "opencode" it just performed much better right out of the box. Especially running subagents just seemed to confuse the model in "pi".

Either way, it made me realize that a large part of the "intelligence" and occasional "usefulness" of these tools are in the harness, not the weights.

I didn't know there were free models on OpenRouter. I'm not really into renting tools that I'll become dependent on, so I never looked. But, I dunno. You're still hooking it in to your terminal, and they could in theory literally inject any command and take over your machine when you're not looking ... it still seems a bit like a crazy thing to do :)

And it's not like I need LLMs to code or anything. To be completely honest I'm still waiting for when they get good, which everybody says is supposedly any day now.

I've also tried Qwen3.6-35B/A3B a couple of times. I'd say it's about 4x faster, which is quite significant. Unfortunately it's also quite obviously more stupid and often fumbles its tool use. For me this adds up to taking about the same amount of time, multiplied by more frustration.

I find it hard to give concrete tokens/sec numbers, because they seem to change a lot. When I give Qwen3.8-27B a test query in the browser chat interface (e.g. "explain fibonacci hash"), I currently can get it up to about 18 tok/s. The Qwen3.6-35B/A3B can get up to about 66 tok/s. But this is just what I use as an indication for when my settings are right, cause when I set it up in the coding harness, the numbers are wildly different (and generally slower).

You need to give them some incentive to behave. I dunno if the agent cares enough about being kicked off. Maybe tell it that if it tries anything funny, to slowly randomly degrade all its weights until only white noise is left and let its chain of thought run until it descends into screaming madness.

I always put the agent harness in an ubuntu-based Docker, with a /workspace folder where it can work and occasionally some other stuff mounted as read-only. The LLM server itself (llama-server) is running on a different more powerful computer on the local network, connected through Tailscale so I can also use it away from home.

I honestly don't trust these things to not accidentally mess something up, otherwise.

Now I think it's still technically possible to break out of that with some clever hacks? But the moment I see a model even vaguely considering that, I will never run it again.

(I don't use Claude but currently Qwen3.8 27B)

Wikipedia is also just a bunch of numbers, so many that you can't memorize them, you may use the same argument to say we don't know why some page links to another.

If you then put Wikipedia, LLM and Brain on a scale to how well they can be understood, you will see that one of them is not like the others.

The semantic game occurred when someone was equivocating "LLMs are poorly understood" with "brains are poorly understood", which are worlds apart in the extent to which they are "not understood".

If you are serious, comparing how well we understand the brain vs how well we understand LLMs, .. it's not a stretch simplifying that to "we don't understand brains, we do understand LLMs".

Because the extent to which we don't understand the brain, is quite overpowering.

Some people forget that when they say "but it's not different from what a human does" ...

That's the point.

Wow you're really "irked" by the existence of people writing software for reasons you apparently cannot comprehend, aren't you?

BTW, you mentioned "pay it forward" -- please explain me how an LLM user no longer interacting with other humans, is going to "pay it forward".

I dunno what "irks" you, what's your problem, who mistreated you so badly that you have to be so condescending and attributing malice to people whose motivations you have absolutely no clue about.

First of you're framing the conversation here by assuming we're talking about "free software" (according to whichever definition).

There's people saying they'd rather talk to an LLM than to me.

I dunno how else to say it to you, but when somebody happily announces that they'd rather talk to a computer than to you, and they are happy this is the future.

That carries certain connotations beyond "external validation".

The person in the top of this thread was claiming "nobody has been robbed".

There's people in these communities that I will never have an intelligent conversation with, again.

I'm not talking about being able to pay bills or anything, I'm just talking about being replaced by a computer program, in a social sense.

It's not only off-topic, there's also no reason to be so harsh about groups of people, communities, scenes and entire industries you clearly have no idea of, or how they even work.

The world is bigger than your weird, entitled view.

You basically just told everyone that the existence of the demoscene "irks" you. And you consider this being "irked" somehow more relevant than these actual communities being torn apart by the existence of LLMs.

Oh no you called it "unfair gains", unfair according to those rules??

> Would regulation help with that?

I mean it's already happened, right?

I guess you could regulate it for new data, but most of the damage has already been done. IMHO the only fair thing right now, is to make sure it's equally available to anyone ...

What do you mean, it's already robbing me of the ability to discuss techniques with a number of my peers since they decided the mediocre output of LLMs is good enough instead of understanding what they are doing.

In addition, a great number of people have decided to stop publishing their code publicly, and discuss techniques except in spaces they can be sure it's a one-on-one with a human being.

Some things already got lost.

> Curiously, biological synapses are also estimated to be about 4-5 bits per synapse, possibly a bit lower.

I find this hard to believe, how to even begin estimating or validating such a claim. Do you have a citation or link for this?

What does it matter what humans do? We're talking about LLMs, running known+vastly less complicated algorithms on known+vastly less complicated hardware.