Nathan🇨🇦

This is probably a “top 5 most brazenly corrupt thing a president has ever done” moment right here. It’s stunning. Sue the government that you’re in charge of, order them to settle, then plunder ~$2B worth of taxpayer money to pay himself and his corrupt idiot friends.

@HighyieldHarry lol. so good

Asked Claude for a company’s customer support number

@TheHitMan1776 @ayeejuju honestly I think this relationship is over and it is her fault. She made a mistake she can't take back

@ayeejuju I understand his feelings are hurt. But dude…sleeping on the couch is a trigger you don’t pull over Legos. Bruh be firing all his ammunition at once.

@Rahll @EpicArtFail You can do this with people too

@NotThatNATO @EpicArtFail You can literally trick it by poisoning the data with a relatively small amount of images, it doesn't 'know' anything. arxiv.org/abs/2310.13828

AI folks love to pretend it was gatekeepers that prevented them from pursuing art, like someone stepped into their house and slapped the pencil out of their hand and blocked them from YouTube, because it's easier than admitting you have zero passion for actually creating things.

@PierceLilholt This seems like the least concerning AI fear of them all. At least to me personally

What happens to trust when AI fabricates perfect relationships indistinguishable from real ones?

@kevinbparry pretty good but I see the cut

Testing a new mirror trick

@IAmSandroSaric @asaio87 and yet I still strongly suspect these programmers are using AI in some form or another, even if its just to summarize documentation

@asaio87 yes, some companies dont allow using AI within their codebase cause of privacy concerns thats legit

I heard there are still companies that have people coming into the office and writing code all day, every day and not use AI too much.

@DeRonin_ @web3_tech_ Before I started doing this, it would burn through the whole days usage on a single request as it would spend 15 minutes taking in everything before formulating a response. One time it wrote the whole feauture and then a python test environment to verify before outputting to me

@DeRonin_ @web3_tech_ Is it usage limits thats causing issues? I'm a pro user and I recently changed my approach. My project code base is getting very large so I explicit tell my claude instance to only review the design doc and the files I reference that are relevant to the feature shipping

Anyone here cancelled their Claude Code and Claude subscriptions and moved to Codex? I feel like I can’t ship the products via Claude anymore…

@EpicArtFail @Rahll You say it doesn't "know" but how can you verify that when it accurately identifies a picture that contains a dog every single time? The system contains that knowledge and that knowledge was learned through trial and error

@NotThatNATO @Rahll it doesn't -know- what a dog is, or even what a picture of a dog is, it has a RESPONSE to something that pings the relative feedback of what the images of dogs it contains are. Its data driven pattern matching.

@EpicArtFail @Rahll It knows what a picture of a dog is. It didn't before. Futhermore I could feed it picutures of anything, or sounds, or smells with the right sensors and get the same results. Its a general learning system. Thats the point

@NotThatNATO @Rahll the pictures and the processing methodology are the instructions and the goal. its all one thing. Even YanleeKun had to circle back on this shit even though he's just been copying Garry Marcus for decades. it doesnt know what a dog is. just the images.

@ProofofMaro Can you explain to me what makes a human sentient? What are the ingridents? What's the formula? Our brains are also very complicated predictive engines Seems to me if we can't define this how can we accurately say a digital advance predictive machine can't be sentient?

Calling glorified auto complete sentient is wild

@EpicArtFail @Rahll no we didn't. We gave it a million pictures. asked it if it was a dog and told it yes you're right or no you're wrong Thats machine learning with labelled data. No programmer defined what a dog looks like before hand, the machine figured it out through trial and error

@NotThatNATO @Rahll machine didnt gain anything new. we gave it the instructions.

@Variety Any real artist welcome new tools that improve their workflow. It's not complicated. These people worked years on perfecting their craft and it's become their entire identity. It's understandable why they'd feel threatened

Jack Antonoff slammed people who use AI to make art and music as "godless whores": "So to everyone who is gassed up about the new ways you can fake making art, by all means drive right off that cliff. We're genuinely happy to see you go. Generations coming will be engaging in the ancient ritual of writing, recording and performing as it comes to us from God." variety.com/2026/music/new…

@Rahll @EpicArtFail I'm saying the result is the same. A new capability is gained by the system (human or machine). You're arguing they're different because the process is different. Is it? How do you know?

@NotThatNATO @EpicArtFail "If not, how can you really compare the two?" You're the one saying they're effectively the same, lol.

@EpicArtFail @Rahll Thats how it feels. But whats happening? What are the underlying processes that make this happen? And if I try to paint the mona lisa a few thosand times I'll make a pretty good representation

@NotThatNATO @Rahll You experience. its an individualistic process where the memory of experience happens. Ever have a thought where you only remember it in the last time you thought about it? You can look at every millimeter of the mona lisa and never know how to recreate it. its not data.

@EpicArtFail @Rahll you have a good understanding how our brain cells capture memories? Pretty hard to argue thats not data compression

@NotThatNATO @Rahll When you read a book, -you- interpret what its talking about, when you view an image - you- take something away from it, or, nothing, its entirely independent. What its NOT is data compression. Its the EXPERIENCE of doing and interacting with something, not the something itself

@EpicArtFail @Rahll 'Learning is the process of acquiring new knowledge, skills, behaviors, or attitudes through study, instruction, or experience. It represents a relatively permanent change in a person's capability or behavioral tendencies...' Impossible to argue machine learning doesn't fit here

@NotThatNATO @Rahll Well now you have Gaining and Developing and have yet to show an Understanding of what learning is as a process. hint* its not data compression.

@EpicArtFail @Rahll Can you easily explain the process of a human brain learning a new skill? If not, how can you really compare the two? You don't understand one of them at all and neither do I. In fact noone really does, the brain is still a huge mystery

@NotThatNATO @Rahll they do not learn. there is no understanding, most of the way the data is processed is humans auditing the correlations made by other humans. The processing is a rule-set packing overlapping details and reproducibility within acceptable loss.

@KaiXCreator it's been working well for me but approach matters. I start new instances regularly. I keep an up-to-date design document that I feed into the new instance. I break every feature into tiny chunks, prompt, test thoroughly and then iterate thoughtfully

Is Claude Opus really the best model for coding right now ?