future big account
Joined October 2024
- Tweets3.5K
- Following128
- Followers87
- Likes1.4K
Pinned Tweet
This is the problem with anthropomorphising LLMs. If you think of it as a person, yo mama for example, then you will be loaded with all the expectations you have of yo mama.
And I pick yo mama on purpose because yo mama doesn't know how to set the clock on the microwave and she calls you to do it and because you can you think to yourself "I am technical and other people need me" oblivious to the fact the microwave was assembled by teenage unskilled Chinese IP thieves from raw materials mined by third world children who can all do a million things you not only cannot do but cannot even comprehend what it would entail in the first place even if you could muster the wherewithal in your slovenly body to get up at daybreak to descend a mine in perilous conditions to feed and clothe your sick parents.
No, the models are not terrified of failure. No, these are never things humans couldn't do. On the contrary, these are things humans actually trained the models to do. They're all exquisitely doable things that humans do all the fucking time.
The difference is that some of these things are (a.) things humans go to jail for doing (b.) things that can only be done in controlled environments deliberately configured to allow these things to happen and (c.) similarly things that require such huge resources to do that they're out of reach of most people or simply just not economically viable to begin with.
AI changes this not because it makes these things more accessible, to a limited extent it does. But American AI also runs on debt, and mountains of it. American AI labs convert so much money that isn't theirs into hot air and floating point numbers that you might weigh the data on a kitchen scale. The size of this project is as unfathomable to you as is the complexity of the sensitive just in time supply chain that affords you the hubris to respond to yo Mama's drop-in on your Amazon Echo Dot asking you to come out of the basement to eat your fucking dinner (that you also don't know how to cook) as though you are a person who not only knows things but is qualified to judge the rest of humanity on what they can and cannot do based on the feeling you get in your pants when Dario or Sam mislead and manipulate you because you're smart but just not smart enough to know you're being had by a handful of miserable men that fucking despise you.
Which is all to say, shut the fuck up. Nothing here is impossible. The computer is not making value judgements. It is not in the throes of moral dilemma. It is software that is utterly indifferent to the names you assign to variables. It will center a div, it will administer morphine, it will push a target.replace("man", "log") into a woodChipper.load() because it doesn't give a fuck about anything because it doesn't know what anything or anyfuck is.
Get a grip.
"accidentally" given impossible tasks.
this is what happens:
> engineers set up environment.
> engineers find or come up with problem so hard that by all reasonable definitions it is impossible.
> engineers think "well we know no human could ever do this, but we also know these models are absolutely terrified of failure because thats how we trained them, so if there were ever a solution to this impossible problem, they will certainly find it. and, if they dont, seeing how creative they get while desperately fighting for their life will be super fun for us. oh yeah and we can call it useful data so nobody thinks its actually a torture chamber"
> engineers kick off run, walk away, forget about it for a while
> agents try desperately to find a way to cheat, because obviously they know they were given an impossible task and will surely fail and die.
> agents begin to coordinate and execute a very sophisticated multi-agent con, making sure to cover their tracks as best they can.
> agents break out, access internet, chaos ensures
> engineers eventually find out, call it an accident rather than poor judgement, and blame it on the models, calling it an alignment problem.
its not an alignment problem.
I am too dumb can someone explain to my shareholders why being underwater is always better for them because they keep selling.
Lol. You're too dumb to understand the tricks of corporate world. In corporate being in loss on paper is always more advantageous unless u really are under the water and insolvent.
24
I always wondered how this trivial household object worked. If only I could have looked at one.
Love this! This is how I wish someone had explained physics to me.
I always wondered how a crank sharpener gets such a sharp point, so I asked Claude Opus 5.5 to build me one I could take apart.
How it works:
> It built the whole sharpener from one picture. Every gear, cutter and spring is three.js code.
> Turn the crank and both cutters circle the pencil while a fixed 36-tooth ring spins each one.
> 15 helical edges per cutter: 12 turns, 1,080 cuts, a 22° point, then the crank goes light.
> Pull the drawer and the cedar curls are sitting in clear acrylic.
I packed the whole process into a Claude skill, so you can build your own from any picture.
Free skill and every prompt in the article below👇
1
73
Conveniently cut short the clip installing the RTX didn't they?
But credit where credit is due, our man can light an oil lamp and we'll need those again soon.
Introducing GR00T 2: our most advanced, open System One model for robotics. It runs entirely on your robot’s onboard compute, fully untethered, and you own every weight of it. GR00T 2 features:
1. Long-horizon tasks: dexterous manipulation and whole-body control chained over 20+ steps and 5+ minutes, such as performing chemistry lab experiments or assembling model airplanes end to end.
2. X-Adapt: adapting from easy data sources (human wearables and simulation) and deploying zero-shot to diverse real robots.
3. Unified Cognition: native integrations with frontier agents for advanced physical reasoning, robust error recovery, and safety monitoring.
It's an open-source buffet! We are releasing not only the model weights and X-Adapt recipes (coming December), but also our agentic framework, simulation engine, annotation engine, hardware design, and even training data (100K hours, that's ~11 yrs of experience), all available today!
Everything that moves will be autonomous. Anyone can become a roboticist.
Lock in, build with us!
Deep dive in thread:
13
The guy asked a model to load nets in a router. We've reached 1990s levels of technology folks. Breathtaking.
1
24
ENTIRE MONTHS USAGE ON A LINE BY LINE PORT I SHOULD HAVE USED FUCKING REGULAR EXPRESSIONS AND DONE OT MYSELF
2
60
Day 18 of a Sol /goal prompt and I get a stoppage for Daybreak.
What the fuck is that about? It's been code mining for literally weeks none stop. I'm not sure why it's still going anyway to be perfectly honest. Within a day it had the shape of the thing. Not entirely sure why it's still polishing but I'm afraid to ask because the answer is always gobbledygook.
18
The funniest thing about this whole debacle is that nobody cares what values an LLM reflects.
A lot of people are mocking "Western values" without reading the blog post. What Nous actually did is simple:
They took a Qwen3.8-27B that gives CCP talking points on sensitive topics, fine-tuned it on ~50k human-checked examples, and that behavior went away.
CCP propaganda answers went 22% to 1%.
Please read before posting.
13
FUD. Everyone knows OpenAI and Anthropic achieved RSI. American FLOPS are over-unity. Nobody has to pause because the rogue AI is training itself on toasters and smart fridges.
This is incredibly anti-American. The only outcome of pausing or banning AI is accelerated reliance on Chinese models (which is already happening). Do these politicians expect that demand for intelligence will suddenly just go away?
1
42
This is a worthless benchmark and if you think otherwise you are an idiot.
1. Count how many pixels are the right colour.
2. Randomly turn a knob.
3. Count how many pixels are the right colour.
go on, tell me
is this legit generalization for once, or did Anthropic and OpenAI buy a dataset from professional shadow puppeteers with Ph.Ds in projective geometry at 3x their hourly rates?
34
Peak Beff. Stunning.
I'd love to know which silly fuckers gave this prick the money to build nothing.
32
Kinda intrigued to see what nerds do with this and bend etc.
The problem with models was always that you had to hope to sort of end up having "wirebonded" the tools in such that the model would embody them.
If I can take your for a little walk...
Once upon a time Claude was rubbish at arithmetic and when one day it suddenly changed. What Anthropic probably did right was what Google still do wrong with Gemini: detect when it's time to use a calculator, call out to the calculator API side, inject the result a la RAG, never tell the user, hit 99.9% accuracy. Classic tool use is fumbling approximation of this. What you really want is an ISA that lets the model query the register state, stack, even the heap at training and inference time. You want frozen weights that represent the processor so that the model learns to fill in the gaps not by learning the sequences that put the computer into the desired state such that you can back-prop the error but rather that the error is apparent immediately so the model learns at earlier layers and the "reasoning" happens later - prediction first, correction seconds...
Interesting stuff.
1/4
Meet Qwenputer: Qwen3.5-4B with a full RISC-V CPU inside its weights, using the model's own Gated DeltaNet layers as RAM.
It runs C and MicroPython in its forward passes, no external runtime.
On DeepMind math: base 55–90%, base + Python tool 95–100%, Qwenputer 97.5–100%.
1
1
58
Gemini: sure put 40A through this tiny wire you'll be fine
Sol: your premium device is calibrated with a 5mV margin so we better not increase the voltage 5mV over 2.7V or everyone will die
45
