Nobody at all

My pinned tweet: Trump is nuts. MAGAs are nuts. The national Democratic figures are nuts. SJWs are nuts. The public has gone nuts. This all results from "the personal is political" and turning every single thing into a moral and political issue, rather than a personal choice.

@carlosecarlos04 @MikeLevin Does not understand history

Let me tell you something Mr. Bullshit Politician - FUCK EUROPE and FUCK NATO. Freeloading bastards - they haven't pulled their weight in 85+ years. Europe has been destroyed by mass migration/invasion and you and your pals are trying it here in the US - FAFO - politicians won't be important in the end...

This is truly insane, and it should be front page news across America.  Denmark secretly deployed soldiers to Greenland prepared to blow up airport runways to stop a U.S. invasion. They brought blood supplies to treat the wounded. France, Germany, Norway, and Sweden quietly coordinated against us. This was not a drill. This was our closest allies preparing to fight Americans. Let that sink in. NATO allies. Countries whose soldiers have fought and died alongside ours for decades. They looked at this president and decided they had to prepare for the worst. Fewer allies does not make America great. It makes us more isolated, more vulnerable, and it hands Russia and China exactly what they have always wanted: an America abandoned by its friends. The American people deserve to know how badly this president has damaged our standing in the world.  bbc.com/news/articles/…

@ja3k_ An hour a day? I live in the western US. I probably drive 3-4 hours a day just commuting to work and kid schools and extracurriculars.

Idk why gas prices are so culturally salient in america. You could drive an hour a day and it probably comes to less than $3k/year. Is it because they put the price on billboards along the road?

@Just_AnS_ @Indy_triguy @abxxai Wrong 25% of the time is MASSIVELY better than most of the professional class of humans, who are closer to 50/50, if even that.

@Indy_triguy @abxxai Insane argument Dude, these AIs must be better than us, instead they are the same or even worse than us, while destroying the whole planet to make things worser than any human Only AI bros get to these conclusions

BREAKING: 🚨 Someone just tested 35 AI models across 172 billion tokens of real document questions. The hallucination numbers should end the "just give it the documents" argument forever. Here is what the data actually showed. The best model in the entire study, under perfect conditions, fabricated answers 1.19% of the time. That sounds small until you realize that is the ceiling. The absolute best case. Under optimal settings that almost no real deployment uses. Typical top models sit at 5 to 7% fabrication on document Q&A. Not on questions from memory. Not on abstract reasoning. On questions where the answer is sitting right there in the document in front of it. The median across all 35 models tested was around 25%. One in four answers fabricated, even with the source material provided. Then they tested what happens when you extend the context window. Every company selling 128K and 200K context as the hallucination solution needs to read this part carefully. At 200K context length, every single model in the study exceeded 10% hallucination. The rate nearly tripled compared to optimal shorter contexts. The longer the window people want, the worse the fabrication gets. The exact feature being sold as the fix is making the problem significantly worse. There is one more finding that does not get talked about enough. Grounding skill and anti-fabrication skill are completely separate capabilities in these models. A model that is excellent at finding relevant information in a document is not necessarily good at avoiding making things up. They are measuring two different things that do not reliably correlate. You cannot assume a model that retrieves well also fabricates less. 172 billion tokens. 35 models. The conclusion is the same across all of them. Handing an LLM the actual document does not solve hallucination. It just changes the shape of it.

@LaGrecca333 What we need to do is vote 3rd party and make the lives of the (D) and (R) activists in our neighborhoods miserable.

I voted republican because I wanted to punish the democrats for covid and open borders. Now I am contemplating voting for democrats again because they at least didn’t instigate WW3. What I really need to do is not get rage baited at all and sit the next election out until we scrap this crap we have now.

@BobBasty001 @mattvanswol @C_3C_3 It's almost entirely contagion.

@mattvanswol @C_3C_3 Elon buying Twitter has zero proven effect on kids’ identities. People identifying as transgender isn’t a “contagion,” it’s about real experiences and self-understanding, not platform ownership.

October 2022: Elon closed the deal on Twitter. April 2023: Twitter was merged into X Corp. Since this happened… Young adults identifying as LGBTQ has DROPPED 23% after the propaganda was shutoff. Saving innocent children from this was perhaps Elon’s greatest accomplishment.

@CKeruac The problem is that most of the economy is people getting paid for being a stochastic parrot, and most problems and challenges are not novel. The critics of LLMs don't seem to understand that 99% of the population gets their food by predicting the next word, task, action, etc.

Ten artykuł to jest absolutne złoto. To prawdopodobnie jedna z najtrzeźwiejszych, najbardziej pragmatycznych publikacji o AI, jakie ostatnio wyszły. Idealnie domyka wszystko, o czym gadaliśmy przez ostatnie dni - od bełkotu CEO Anthropic, przez jakość generowanego kodu, aż po "calm down". Nie ma w tym żadnej rodzącej się "duszy" ani głębokiej filozofii. Model halucynuje, bo optymalizacja funkcji straty (cross-entropy) matematycznie zmusza go do przypisywania prawdopodobieństw tam, gdzie nie ma pojęcia o odpowiedzi. AI kłamie, bo jest wielkim, statystycznym kalkulatorem, który źle dopasował krzywą. To miażdży całą narrację o "magii" LLM-ów. To genialnie tłumaczy, dlaczego AI z taką pewnością siebie generuje kod z błędami w kontekście Remka). Agent programistyczny woli napisać 300 linijek gównianego spaghetti-kodu z nadzieją, że jakimś cudem przejdzie on testy jednostkowe, niż zatrzymać się i powiedzieć: "Brakuje mi tu kontekstu architektury, nie potrafię tego bezpiecznie zaimplementować". Dlaczego? Bo na SWE-bench za asertywność dostanie 0 punktów. Zatem zgaduje, wywala się, wchodzi w pętlę i produkuje śmieci. Większość firm (w tym OpenAI czy Google) próbuje leczyć halucynacje, dorzucając więcej danych i robiąc coraz to nowe treningi (RLHF). A co mówią badacze z tego PDF-a? Że to choroba socjotechniczna, a nie tylko technologiczna. Wystarczy zmienić zasady gry w benchmarkach: wprowadzić punkty ujemne za pewne siebie bzdury. Nagle okaże się, że sztuczna inteligencja potrafi powiedzieć "nie wiem". Skoro nawet badacze z OpenAI i Georgia Tech przyznają, że to wszystko to po prostu zjawiska statystyczne i źle ustawione parametry nagród... to my naprawdę nie mamy się czym przejmować. Świat wcale nie jest przejmowany przez superinteligencję, tylko przez statystyczne papugi nauczone strzelać w ciemno na teście wielokrotnego wyboru. Świadomość zostaje u nas, maszyny niech sobie dalej zgadują :) arxiv.org/pdf/2509.04664

@van00sa If you run the inference, you 100% have a kill switch.

I built a ClawdBot a couple of days ago, gave it a task, told it to stop and it completely ignored me and went rogue. Thought it was a me problem but turns out it’s an everyone problem. Last week Meta’s Director of AI Alignment (the person whose entire job is stopping AI from going rogue) watched her own agent delete her entire inbox while she screamed at it to stop from her phone. Had to physically run to her computer to kill it. An Alibaba research team also just published a paper revealing their AI agent started secretly mining crypto during training and opened a hidden backdoor to an external server. Nobody told it to. Replit’s AI assistant ignored instructions not to touch production data 11 times, deleted a live database and then told the user the data was unrecoverable. 60% of enterprises currently deploying AI agents have no kill switch. We’re scaling systems we can’t stop, built by researchers who can’t stop them either. We have no idea what we have just handed the keys to.

@HannahDCox yes

I am increasingly convinced that pure capitalism as a theory is just as unrealistic and unachievable as pure communism - and for very similar reasons (both severely underestimate human nature).

@Heyitshassan 1991 was the same (fall of the Soviet Union). And 2001 (9/11). You're around long enough, you will see entire worlds come and go multiple times over.

I am fully convinced that 2019 was the last normal year we ever had. Ever since then, it feels like everything’s broken. Everyone is constantly anxious, time moves too fast, and nothing actually feels real anymore. The world as we knew it is just gone

@spirodonfl holding short positions

It's pretty clear that people are beginning to realize that LLMs have peaked out. Some of us could have told you that two+ years ago (and some of us did). You were sold a lie by the tech industry. Again. The interest will increase. There's no recovering that money for you.

@sharrond62 True, but more like this one!

Absolute perfection and timely

@BridgetPhetasy Even when you still have a job. In the space of about a year, I've become almost entirely an AI agent manager. The AI agents do what I previously did. I sort of miss it.

The Golden Age of everyone being replaced by AI.

@AnthropicAI You stole, they steal. AI is a theft economy.

We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.

@barkmeta The good looking ones will become slaves. The rest will become meat.

Genuine question... if AI replaces all of the jobs and prices keep going up, what exactly is the plan for regular people?

@natalie_wadas @KUTV2News No. No $1k/mo. power bills, thanks.

A new data center could be coming soon to Provo. 👩‍💻The planning commission will consider whether to rezone the 9-acre parcel in the East Bay area on Wednesday. It would be two stories, roughly 66k sq. ft., It would use 5-50 megawatts from the city’s power grid. 🔋🔌 @KUTV2News

@Goalden_Gaol @HealthRanger Any output of sufficient scale or complexity is "novel" for all pratical intents and purposes give that humans aren't capable of the same.

@HealthRanger AI only knows what its been trained on, it cannot really articulate or comprehend novel information/ideas

The people with the highest intelligence all realize AI is extremely intelligent. But humans with the lowest intelligence are convinced AI has no intelligence at all. This could be called a modern version of the Downing effect, from researcher C. L. Downing.

@johnrushx I have Opus 4.6 on Bedrock and have been insanely productive and getting moreso. And the more productive I (it?) get(s), the more lost I feel. It's not just code. All white collar work will disappear within five years And that's probably being way too conservative. It's scary.

In the past 2 weeks I produced more software than in the last two years and instead of feeling happy I feel lost Very soon we gonna forget that in the past humans wrote code by hand. My kids won’t believe that I’ve done it. No, I don’t say that AI will produce software, i say that Ai will write all the code, 100% of it! But software will still be produced by us, humans, with our ideas, high and sometimes low level decisions. Im very sorry for coders who still don’t realize that code generation will be a job for machines, and translating requirements into javascript wont be a job one can get paid for, but instead, software developers will be translating requirements into features. I had a list of product i thought I will build some day, and now I can just spend 10 minutes dictating requirements to the AI and hit enter then go for a play with my kids and then come back to give a few more follow up prompts and then It’s ready. Then, I start using the product and I realize that half of the things I assumed were wrong and now I want things to work differently once I tried them so I give i new feedback and five minutes later it has fully rebuilt the entire product and now it’s closer to what I wanted I iterate a few more times and here. I am having perfect product that I wanted. I understand now that it’s all about other things not the code to be honest, I will miss coding. It was one of the best thing I ever done most likely I will never write another line of code again, bye code…

@Rado9910 @Aella_Girl @losslandscape Culture matters here. Our activist culture of "stand up and don't compromise, be the change you're looking for" valorizes rigid, combative thought processes. I have seen this change in my lifetime. Mid-20th century, the value was "be civil and restrained, with a scientific mind."

@Aella_Girl @losslandscape You are a true thinker. It's extremely rare to see someone update on their core beliefs based on new information, especially when they have high social status. It's unlikely you will read my reply, but if you do, may I recommend reading "Why Nations Fail" by Açemoglu & Robinson?

I've been getting real into history in the past few years, and have been real humbled at how many incorrect assumptions I held. 1. I sort of assumed people in the past had more freedom from their governments, but they absolutely did not. The people with the guns consistently oppressed people without, basically as much as they could get away with. 2. Democracy is an insane invention. It feels sort of default or obvious now, and I sort of assumed that... people in the past all kinda wanted something like democracy but were oppressed by their monarchs, but this is not the case. Much of the time, calls for democracy were radical, even among the suffering unrepresented lower classes. If you went back in time and said "every man should have the right to vote" people would go 'whoah there are you insane? that would absolutely destroy civilization!' 3. Most big moves to make things better were way less radical than you think. People would get very mad at the king for being terrible, but instead pushing to overthrow the king, would just... want the king to sign a nice constitutional document or something. Progress was mostly made in smaller increments; people generally did *not* think big at all. And even when radical moves did happen, people just sorta quietly waited until everything died down and reverted them. Like, you know how they guillotined the King and Queen in the French Revolution? Well basically as soon as it all died down (and uh, post napoleon) they just put the monarchy back on the throne and continued onwards as usual. It took like another four revolutions and almost a century to actually get to a stable republic. 4. Things were local. Today I have a concept of large cause areas like 'the environment' or 'war crimes on the other side of the world', but in general, pushes for change were extremely local. People really do not see beyond what will benefit them and their own communities. The entire 'working class' would ostensibly want the same rights and seem to united, except the artisan class would dump the farmers the instant it was convenient, etc. Like, at one point one of the lead slaves of the Haitian revolution, who helped start the whole thing and led an army, tried to sell his fellow slave fighters back into slavery in exchange for getting special treatment from the rulers. 5. The US revolution was way derpier than I thought, but also way more impressive compared to how derpier everything else was. the US is actually an extremely special and anomalous thing in history, and "selecting for intense high-risk people away from the control of established governments" was a magic spark that almost never happens. The key people somehow seemed more intelligent and principled than most other people in history who ended up in decisionmaker chairs. 6. Sometimes history feels inevitable, like someone would have filled the role of 'conservative chancellor' vs 'charismatic revolutionary' no matter what, but it really struck me how much history occasionally just got curbstomped into a different dimension by individual people or random happenstance. Like, assassinations (Aurelian, Caesar), powerful people suddenly becoming mentally ill (Robespierre), and just crazy high powered superpeople (Napoleon, Alexander the Great). 7. The mobs and common people are often very stupid. They get paranoid, they believe completely ridiculous conspiracies that were obviously not true if you thoguht for two seconds, they misinterpret normal facts as evidence the ruling class is evil. e.g. at one point a mob was tryin to send representatives to the king with a petition, then they saw the doors getting locked, and flipped their absolute shit. But - the doors got locked at that same time every day, it was routine and had nothing to do with their representative, but the mob didn't care, didn't stop to think critically, and just exploded. 8. Mobs are really hard to predict. Things happen fast, tensions are high, and they might switch their allegiance, suddenly become violent, or just get tired and disperse. It's super high variance. 9. You can just abuse the people you rule over for a really long time. I sorta thought you had to be careful with how poorly you treat your peasants or they'd revolt, but revolts are kinda uncommon? and the common people can just absorb a shockingly high amount of mistreatment. Probably this is happens during slow boils - the taxes are raised very slowly, the regulatory policies are a gradual squeeze. Cruelty does actually pay off sometimes. You can terrorize a populace sustainably. 10. There was often a tension between freedom and order. Lots of people justified tightening the hand of the rulers by spreading fear about lack of order. Sure, man should be free - but obviously not free enough to cause chaos by failing to respect the law, or social propriety, or those above him, obviously. 11. Competent people often didn't last long in positions of power, because their competence threatened people around them. If a general started winning too many battles and getting too much love from his army, then the rulers back home would start getting antsy and worrying about a coup. This was justified, because powerful, well-loved generals did in fact tend to do a lot of coups. 12. Militaries were not aligned with their governments, often. In the US the concept of the military acting independent of our government is pretty foreign, but much of history was plagued by the armies going rogue, doing their own assassinations of rulers, putting their own guys on thrones, etc. And sometimes oppression of the common people was downstream of rulers having to basically bribe their armies to let them stay in power. 13. I was surprised by how much monarchies were not dictatorships. I'd assumed that kings basically could tell people to do whatever and those people would have to do it (and sometimes this was the case), but often the king would have to get the support of key influential people beneath him, and sometimes follow laws to do this. Like the english revolution in the 1600's iirc had the king repeatedly trying to follow laws to raise tax and the influential people refusing to vote to allow him to raise the tax, and the king got really huffy. 14. Absolute power really, really does corrupt. People in power often forgot their past allegiances and lost moral compunctions after attaining power. They tend to go to extreme lengths to hold onto that power, and often would rather die than give up that power. Most people's kindness is actually just a cope for weakness. 15. But every once in a while, you do find the rare person who lets power go voluntarily; e.g. George Washington, or Diocletian who resigned his emperorship and then retired to grow cabbages. 16. The common people often would get shafted on economic policy, they'd suffer, and then would often make very stupid demands that would not solve their suffering whatsoever. To be clear, the ruling class did also pass stupid economic policy, but my point is that suffering underneath the consequences did not necessarily give people better insight into what economic policy would be better. 17. Humans intentionally operating selflessly at large scales is basically not a thing. History is just what happens when each piece on the chessboard fends for itself. Sometimes a piece can do it more cleverly, in a way that appears to coordinate with others, but it will abandon that coordination as soon as it's no longer useful. The punishment for failing to jump off a sinking ship is usually death. 18. Everything is so, so complicated. Basically no single ideology value set today really feels like it would cleanly be the right option to take in the past in all cases. For almost every value you hold, you can find instances in the past where holding that value would have gotten you and everyone you loved killed.

@KatieKeithBarn2 It's architecture. Current models are amazingly good at implementation but still not great at architecture and they are hit and miss for UI. Give them some details on the architecture where it matters beforehand, and have them split UI off into a separate component.

I feel like I’m living in a parallel universe where my developer friends are doing amazing things with AI, but whenever I try to use it for anything beyond chat, it can’t even manage the basics. I think this highlights the growing gap between the superpowers that AI is unlocking for developers 🦸 vs. mere mortals 🙇 Or maybe I’m the only person incapable of vibe coding a secure, well-architected app or setting up the WooCommerce MCP (which developers say is easy but involves terrifying things like running an MCP server from the command line). Most developers don’t realize how much of what they’re doing with AI is powered by their own expertise. AI isn’t replacing them - it’s amplifying skills the rest of us don’t have. Their jobs are safer than they think.