Sabitlenmiş Tweet
Piero Raimondi
25.8K posts
@Glenn_Diesen Game theory rewards collaboration in repeated games:Tit-for-Tat sustains mutual cooperation for higher long-term payoffs. You nail it: Western"hegemonic peace" irrationally treats security as zero-sum,rejecting diplomacy as appeasement instead of harmonizing interests
English
We are told that security in the Middle East requires defeating Iran, security in East Asia requires defeating China, and security in Europe requires defeating Russia. We never discuss security in terms of how to learn to live together by harmonising interests and managing competition. This is by design. This is hegemonic peace, in which security depends on defeating rivals rather than managing a balance of power.
Subsequently, security relies solely on deterrence rather than reassurance; diplomacy is dismissed as appeasement; peace agreements are temporary and deceptive; and war is peace. Our rivals do not have legitimate security concerns, as their policies are allegedly always motivated by aggressive, irrational, or expansionist behaviour.
We have convinced ourselves that our liberal hegemony is a force for good, and that our opponents oppose our dominance because they reject our benign values of freedom. Discussing the security concerns of adversaries is believed to “legitimise” their policies, which is treasonous. The world is divided into good guys (liberal democracies) and bad guys (autocracies). We should not ask how defeating Russia, as the world's largest nuclear power, is a rational security strategy, or why our governments refuse to even speak with Moscow to discuss the European security architecture and end the war. Our governments have relabelled nuclear deterrence as nuclear blackmail to signal that there can be no more constraints.
All empires can become irrational during decline. Leaders take greater risks to avoid decline, legitimacy crises at home must be distracted with enemies abroad, outdated strategies from a bygone era of strength are still embraced, and there is a tendency to double down on narratives of being indispensable, representing universal values, and dismissing all opposition as illegitimate and dangerous. Are we the fanatics?
English
@NousResearch It seems to be the next level of coding agent, but I'm not able to use it with a Z.AI GLM coding plan subscription... It is throwing 429 errors when I choose recent models (i.e. GLM 5), only working if I choose outdated models (GLM 4.5 flash)... Any clue?
English
Don't let your agent get framemogged.
Use Hermes Agent.
English
@pati_marins64 Good point.Just two quick thoughts: the main disagreement was about keeping AI out of mass surveillance of American citizens (that distinction feels important).Second, humans have caused plenty of mistakes long before AI existed, this tech mostly amplifies risks we have I think
English
The world needs to pay attention to this
Over the last few weeks, a major ethical debate has been unfolding in the U.S, and it concerns the very future of warfare.
The conflict involves an unprecedented standoff between the Pentagon and the AI company Anthropic (creators of Claude).
Anthropic held a contract with the Pentagon that included strict "red lines": their AI could not be used for mass surveillance of citizens nor for the development of fully autonomous weapons.
However, with the contract already in effect, the Pentagon demanded the removal of these ethical safeguards. When Anthropic refused, Sec. Pete Hegseth designated the company a "national security supply chain risk," and President Trump ordered all federal agencies to stop using Claude.
The government quickly moved to replace Anthropic's role with OpenAI and xAI.
While these are new contracts, they fulfill the same strategic functions but with a crucial difference: they have reportedly agreed to operate under the Pentagon’s existing policies, specifically Directive 3000.09.
Why is this a turning point? Because Directive 3000.09 does not prohibit fully autonomous lethal weapons.
It allows for "appropriate levels of human judgment," a vague term that the military interprets as allowing AI to trigger force independently once a mission is programmed. If a human programmed the AI to kill, it’s fine.
By accepting these terms, OpenAI and xAI are effectively removing the private ethical barriers that Anthropic fought to maintain.
Anthropic has now sued the U.S. government, alleging illegal retaliation for its ethical stance, a move backed by Microsoft, which warns of a dangerous precedent.
This is about more than just software; it means AI could soon make life-or-death decisions on its own.
We see the errors AI makes every day; these systems are nowhere near ready to decide the fate of a human life.
Global society must keep a close eye on this issue.
English
@rohanpaul_ai @bcherny Just a correction, he says he's coding with Claude Code, not Copilot
English
@AleEquilibrium 🔝
e senza contare i derivati l, es. la pubblicità, che diventerà personalizzata e generata ad hoc, influenzando in maniera mirata le scelte economiche di ciascuno e quindi sociali
Italiano
aumenta negli adolescenti immersi nel mondo virtuale. Una rivoluzione che va studiata (soprattutto per l'impatto sul nostro modo di vivere), dato che l'arte, la musica, i libri, i film, etc., hanno definito profondamente lo sviluppo delle nostre società fino ad ora.
3/3
Italiano
Riassuntino geopolitico del 2025 in stile anime giapponese degli anni '90. La cosa da notare è che questo video è frutto dell'evoluzione dell'intelligenza artificiale degli ultimi mesi. Si parla tanto dell'impatto sul mondo del lavoro, mentre
1/3
Autore del video: @PsyopAnime
Italiano
I sometimes worry that getting a larger profile invites an inevitable backlash where people talk about how terrible you actually are. My genius plan if this happens to me is to do an Alexander Hamilton and release a document detailing all my personal moral failings.
English
@lefthanddraft Of course but all AI is trained on human data...and human feelings
English
Those of you worried about AI takeover: do you ever look at humanity and wonder if you are on the wrong side?
English
A Caracas è esplosa la battaglia per le strade. Ignoti i motivi e ignoti gli schieramenti.
Potrebbe trattarsi di una resa dei conti intestina al sistema bolivariano: fedeli di Maduro contro suoi oppositori.
WarMonitor🇺🇦🇬🇧@WarMonitor3
Video of major gun battles ongoing around the Venezuelan Presidential palace.
Italiano
I want to start a community dedicated to Claude Code.
It’s become the gateway drug to coding and experiencing the power of AI for tons of people.
This will be a space for people to share killer use cases, agentic workflows, proven prompts, and connect with other CC obsessives.
Comment “Claude” if you want to join.
English
@emanuelpietrob1 Argomenti che dovrebbero essere al centro delle discussioni politiche di ogni partito: pensare al futuro della società, in pieno declino demografico /democratico, senza una pianificazione adeguata. Senza una (1) idea del futuro. Temi approfonditi spesso anche da @AleEquilibrium
Italiano
Ha a che fare con la nostra politica di integrazione, che semplicemente non esiste. Indigenti da tutto il mondo che entrano, vengono ammassati nelle periferie e incontreranno più ostacoli che altro nel corso delle loro vite. Uno su mille ce la farà. Modello Francia. /9
Italiano
Noi e l'Islam, una riflessione senza ambizioni.
L'odio è in aumento. Normalizzato, reso fashion e, purtroppo, perfino "istituzionalizzato" da figure che, nel nome dell'interesse nazionale, dovrebbero lavorare alla convivenza pacifica tra le comunità della nostra Nazione. /1
Italiano
@emanuelpietrob1 Buon 2026! E grazie per questa svolta che ho percepito nei s(t)uoi tweet ultimamente, più "umani", perché meno strettamente analitici a tutti i costi... e più emotivi
Italiano
In un mondo che vi chiede di essere dei guerrafondai alla Douglas MacArthur, ossia delle persone che pensano che la guerra sia la soluzione a tutto, siate dei diplomatici alla Dag Hammarskjold che dedicano la vita alla convivenza pacifica tra i popoli.
Buon 2026.
Italiano
@jenzhuscott Claude Code + Opus 4.5. The first time i felt almost useless also about planning 😊
English
Happy Christmas, one and all!! 🤗💕
English
@kimmonismus It seems that Italy’s trend differs. I think that with a high median age (48.2y in 2025) and few young people adopting AI, the gap grows. Needs strong digital education push to catch up...
English
@kimmonismus We don’t even have definitions for "consciousness" yet. I suspect we’ll spend our time tweaking our definitions while technology surpasses our understanding. Ultimately, even Gödel’s incompleteness theorem hasn’t stopped us from discovering black holes or quarks... or AI
English
Yann LeCun argues that human intelligence isn’t truly “general” but sharply specialized and brutally inefficient for most possible computations, even if it’s theoretically Turing-complete.
Like a shallow neural net that *could* approximate any function but only with absurd size, our brains only cover a tiny, structured corner of all possible functions, while most of reality remains entropy we can’t really grasp.
To be fair, I believe we need clearer definitions of what general intelligence, universal intelligence, and especially artificial general intelligence are.
Demis Hassabis@demishassabis
Yann is just plain incorrect here, he’s confusing general intelligence with universal intelligence. Brains are the most exquisite and complex phenomena we know of in the universe (so far), and they are in fact extremely general. Obviously one can’t circumvent the no free lunch theorem so in a practical and finite system there always has to be some degree of specialisation around the target distribution that is being learnt. But the point about generality is that in theory, in the Turing Machine sense, the architecture of such a general system is capable of learning anything computable given enough time and memory (and data), and the human brain (and AI foundation models) are approximate Turing Machines. Finally, with regards to Yann's comments about chess players, it’s amazing that humans could have invented chess in the first place (and all the other aspects of modern civilization from science to 747s!) let alone get as brilliant at it as someone like Magnus. He may not be strictly optimal (after all he has finite memory and limited time to make a decision) but it’s incredible what he and we can do with our brains given they were evolved for hunter gathering.
English
@DmytroKrasun @mrjogoh I think you pointed out why prompting still remains important.Best practices say: provide detailed context (i.e. your plan), what you want to achieve (improve it!), what you don't want to see in the answer (do not overenginner an already good plan)
English
@mrjogoh A few first versions were really good.
Later, it was complete over-engineering.
English
I asked Opus to write a feature plan and then asked GPT to improve it. It found flaws and proposed improvements.
But then I asked Opus to improve the improved plan and again sent the result to GPT.
And they kept finding issues and improving it...
Why doesn't the process converge at some polished version?
English
@rohanpaul_ai LLMs process at light speed, kids crawl. Yet the kid learns causality by touching fire once. Add embodied world models + cheap durable hardware, and that speed finally counts. The gap closes in years, not millennia.
English
Yann LeCun's new interview - explains why LLMs are so limited in terms of real-world intelligence.
Says the biggest LLM is trained on about 30 trillion words, which is roughly 10 to the power 14 bytes of text.
That sounds huge, but a 4 year old who has been awake about 16,000 hours has also taken in about 10 to the power 14 bytes through the eyes alone. So a small child has already seen as much raw data as the largest LLM has read.
But the child’s data is visual, continuous, noisy, and tied to actions: gravity, objects falling, hands grabbing, people moving, cause and effect. From this, the child builds an internal “world model” and intuitive physics, and can learn new tasks like loading a dishwasher from a handful of demonstrations.
LLMs only see disconnected text and are trained just to predict the next token. So they get very good at symbol patterns, exams, and code, but they lack grounded physical understanding, real common sense, and efficient learning from a few messy real-world experiences.
---
From 'Pioneer Works' YT channel (link in comment)
English
@stebaraz Non ci andranno, ci andremo noi o i nostri cari, i nostri figli o nipoti... Loro sorseggeranno champagne piangendo le nostre scomparse... Ma quanti siamo stati eroici! In alto i calici allora!
Il nostro "viaggio al termine della notte" non è mai finito mi sa
Italiano
Spero lui, Calenda e la Kallas in rampa di lancio nel primo assalto.
Clash Report@clashreport
NATO's Rutte: We must be prepared for the scale of war our grandparents and great-grandparents endured.
Italiano