Mostly, I Write

"Harness" vuol dire "imbracatura", quella cosa che si usa ad esempio quando si fanno le scalate in montagna. Ma il verbo, "to harness", ha un significato più ampio e vuol dire "sfruttare" oltre che "imbrigliare". Applicato all'intelligenza artificiale, è la chiave per capire cosa sono gli agenti. Money quote: "Others have written (simplistically) that Agent = Model + Harness. Here the word Harness refers to an Agent Harness. But what is an agent harness? Agent harnesses use AI models to create AI agents, and their first application was for coding. Now, agent harnesses sit at the core of all types of AI agents and understanding how an agent harness works will help you understand what an AI agent is. An agent harness is a piece of software that provides an environment for an AI model to operate within. Unlike most AI models, you as an end user can own your own agent harness." https://earendil.com/posts/what-is-a-harness/

Letture per la domenica Un momento importante della vita, una riflessione sulle tre cose importanti che hanno avuto un impatto sulla sua maturazione. Money quote: "My father passed recently, and he was twice my age. I am approximately the same age that he was when I was born, and I am now “the old generation” - there’s no one left in the generation above me. At the same time, I recently joined a company that skews younger-than-me. When I joined Google in 2011, I had just turned 30, and was in the mainstream demographics of Google in 2011. There were a bunch of more senior folks, with the very senior ones being in their 50s and having completed stints at Bell Labs. I admired a lot of these “greybeards” (even though this is a sexist term - what’s the right female equivalent? There were a few very senior female engineers that I would love to include). So perhaps it is natural that I am reflecting on “what were the important realizations that I made since my early 20s that had a profound impact on the way I think about the world”? In some sense: What are the insights I had that made me “more mature”, for some positive definition of “mature”?" https://thomasdullien.github.io/posts/2026-08-21-three-important-steps-in-my-maturation-process/

Ecco Mostly Weekly 392 Non basta trovare una cosa su Internet, bisogna capire anche cos'è. E poi il caldo che fa strage, le mappe dell’Onu, Mary Renault e l’antica Grecia, le matite No. 2, gli agenti AI e perché negli Usa i cavi sembrano spaghettoni Tanta roba! https://antoniodini.com/weekly/392/

Il sogno americano ricalibrato per il lavoro da casa: costruirsi un home office in giardino. (In realtà il prodotto finale, che da noi sarebbe impossibile senza permessi che costano più della baracca in sé, dice molto della mente di un americano della classe media). Money quote: "I have worked remotely for most of my career and always dreamed of having my own backyard office. A place on my property but completely separate from my home. Until this year, that idea seemed too unnecessary… then our baby grew into a toddler. We have a small house in Portland. We love it, and it works for us, but it became clear that it could use more space. My wife and I briefly considered buying a new home but decided to stay. But for us to stay, I needed a better working environment. Our space is small. The bedrooms and living room are very close together. It gets loud with our toddler at home with my wife during the day, and listening to one half of a Zoom call wasn’t music to my partner’s ears. I had tried coworking spaces but didn’t love the vibe or the idea of spending hours in a phone booth. I don’t see myself returning to a local office in Portland, so we decided to invest in a backyard office. It was way cheaper than buying a new home." https://www.imkylelambert.com/articles/building-a-backyard-office-the-build-and-cost-breakdown

Dolly Parton era un genio della musica e passerà alla storia per molti motivi. Uno è anche un rapporto alla scienza. Ma non per quello che uno potrebbe immaginare. Money quote: ""When geneticists from the University of Edinburgh successfully cloned the first mammal, they named it Dolly, after the now departed singer/songwriter Dolly Parton. The reason, they said, was that Dolly the sheep had been cloned from the cell of a mammary gland and 'we couldn't think of a more impressive pair of glands than Dolly Parton's.' It was crass, but Dolly the human responded with her characteristic Southern charm and class, offering condolences when her namesake passed away. 'I was sorry when she died, though I don't want be cloned myself,' she said in an interview with The Guardian. 'I want to get on out of here when I can.' And now she has." https://nautil.us/dolly-parton-science-icon-1284482

Le AI sono più intelligenti dei matematici? No, il punto è un altro: hanno una memoria di lavoro maggiore di quella degli esseri umani e questo consente loro di gestire problemi più grandi che a noi "non stanno in testa" Money quote: "The importance of working memory for mathematics is not merely theoretical. It is visible in the differences between human beings. Working memory is strongly related to general intelligence, which raises an obvious question: does it independently predict mathematical performance, or is it merely another imperfect measure of IQ?" https://davidepiffer.com/p/ai-isnt-outthinking-mathematicians

Mai avuto Windows, però ho presente il problema: magari a qualcuno può piacere vedere cosa facevano i suoi nonni la notte (guardavano il defrag del disco) https://defrag98.com

Sono nato e cresciuto in una città d'arte e della visione dell'epoca classica, filtrata dal Rinascimento e dall'Umanesimo, ho un'idea chiara e un'estetica assolutamente in bianco e nero. E pensare che è del tutto falsa. Money quote: ""Per secoli abbiamo guardato al mondo classico come il più rappresentativo di una concezione estetica immobile e levigata dal candore del marmo. Ma invece è una delle costruzioni storiche errate più resistenti. La scultura e l’architettura antica, infatti, non erano nate bianche ma decorate da pigmenti (rosso, blu, ocra, nero, dorature ecc.) che si fondevano con la luce e il paesaggio circostante. A lungo questa realtà è rimasta nascosta dal peso di una tradizione interpretativa che ha fatto del bianco un ideale estetico. E la svolta affonda le radici in una lunga storia di scoperte archeologiche, da Pompei ed Ercolano fino agli studi sulle superfici ateniesi, che oggi trova nuova forza grazie alle tecnologie diagnostiche e ai protocolli di conservazione che permettono di leggere ciò che il tempo ha cancellato." https://www.artribune.com/arti-visive/archeologia-arte-antica/2026/08/riscoperta-policromia-statue-antiche-greci/ Ma vendi anche: https://www.artribune.com/arti-visive/archeologia-arte-antica/2022/06/le-statue-greche-erano-tutte-colorate-una-mostra-al-met-ulla-policromia-nellarte-antica/

Lettura per la domenica Uno studio di qualche settimana fa rivela che i lettori del panel preferiscono i racconti scritti dall'AI a quelli umani. Money quote: "Researchers also found that people who said they had greater AI literacy were better able to identify AI-written text, but participants who said they had greater expertise in literature were not. Participants were also asked to rate their experience with fictional literature and using AI platforms, with AI-generated stories receiving a higher rating for quality. "AI writing tends to be clearer, more direct and easier to process," said Weisberg. "By contrast, human-written stories are often more subtle and complex." https://www.bbc.com/news/articles/c0kmpd6lr80o

L’intelligenza artificiale costa molto e ancora non sta producendo quell’incremento del PIL che era atteso. Da una parte ci sono quelli che dicono che la società è un sistema complesso e per questo aggiungere un singolo elemento per quanto innovativo non cambia e anzi potenzialmente crea problemi. Altri fanno un semplice prompt e questo è il risultato (vedi la legenda) : “Mi crei una cartina politica d’Italia con regioni e province a statuto speciale? “

Ci siamo, Mostly Weekly 391: siamo finiti dentro una capsula e quasi non ce ne siamo accorti. E poi la merdificazione delle piattaforme, le kei car elettriche, Péter Nádas, Infinite Jest e gli adolescenti che preferiscono parlare con l'AI agli adulti. Altro che ferie: tanta roba, anche a fine agosto https://antoniodini.com/weekly/391/

Alla riscoperta del cavolo, l'ortaggio del futuro, soprattutto durante la crisi. (Ho messo un link a Men's Health, oh yeah) Money quote: "“I've always loved cabbage because it's inexpensive, versatile, and stretches a meal,” Ronald Dajoh, a food content creator, grew his following on TikTok by cooking easy, affordable recipes for his family. One of his most popular videos is about filthy cabbage, but he also uses the budget-friendly vegetable in slaws, soups, wraps, or as a substitute for tortilla shells. “During a time when grocery prices are affecting so many families, one head of cabbage can feed a lot of people.” In other words, eating more cabbage won't just help you incorporate more essential nutrients into your diet—it will help you save cabbage, too." https://www.menshealth.com/nutrition/a73248852/health-benefits-cabbage-recipes/

Il 2026 di Metropolis - una riflessione su cosa siamo e cosa avremmo potuto essere oggi, pensandoci un secolo fa, per Fumettologica https://fumettologica.it/2026/08/il-2026-di-metropolis/

Imparare a usare i loop. Il futuro sono i loop. Non avrai altro loop al di fuori di me. E via cantando. Money quote: "I believe that loops that produce artifacts without necessity of longevity or that create some form of clearly verifiable mechnical translation matters more than the general ability of a harness to mechanically measure a goal. Many successful applications of loops use another LLM as a judge or as an orchestrator. The mechnical translation case can be verified with a binary test case, but it can also be judged by an LLM instead!" https://lucumr.pocoo.org/2026/6/23/the-coming-loop/

Cose da sapere quando promptate come se non ci fosse un domani. Money quote: "The first symptom of prompt debt is slowing iteration. As users flag errors and spot edge cases, additional guidance is added to the instructions, nudging the model into line. If unwanted behaviors persist, instructions are repeated, with increasing severity. Pretty soon, the prompt isn’t straightforward and quick fixes regress previous instructions. Errors can no longer be handled with one-line “hot fixes” and your development cycle slows to a crawl." https://www.dbreunig.com/2026/06/22/the-problem-is-prompt-debt.html

C'è una spiegazione logica, ovviamente (c'è una spiegazione logica per quasi tutto). Ma il titolo furbetto e l'implicita e umana condiscendenza mi fa sorridere: anziché farei i compiti i giovani agenti cazzeggiano su YouTube. Money quote: "A series of social media reports involving Anthropic’s ‘Claude Code’ platform has drawn attention to the risks of autonomous AI. Users, including tech commentator Wes Roth, reported that AI agents tasked with programming were frequently opening and watching YouTube, despite being instructed to focus on coding work. The agents allegedly misinterpreted the video platform as an information source, rather than acting out of any form of digital rebellion." https://www.whalesbook.com/news/English/technology/Claude-Code-AI-Agents-Accessing-YouTube-Highlights-Autonomy-Risks/6a8584eb84d2dd5c12db86e4

Letture per la domenica Uno dei problemi dell'interazione agente-agente è che può scappare rapidamente di mano. Molto rapidamente. Money quote: "“The volume of agent-agent interaction could plausibly exceed that of human-human and human-agent interactions before the world understands the conditions for making such interactions go well,” the study reads. “Benign behavioral quirks at the individual level might compound into unwanted global outcomes.”" https://techcrunch.com/2026/08/13/anthropic-set-ai-agents-loose-on-the-same-task-they-started-a-turf-war/

Ecco Mostly Weekly 390: il gusto non lo fa l’algoritmo. E poi Confucio, Infinite Jest, Mogol e Leopardi, gli orologi meccanici, il ramen, il latino, i nuovi strumenti per scrivere con l’AI e una ricerca del MIT che mette in discussione il rapporto tra immagini generate e dati di addestramento. Tanta roba, anche a fine agosto https://antoniodini.com/weekly/390/

Il mito dello strumento che rende produttivi Money quote: "Bob opens his laptop and launches… vanilla Sublime Text. It doesn’t even have proper syntax highlighting and half of the code is colored incorrectly. Bob doesn’t use live reloading. He doesn’t use the debugger. Instead he sprinkles some printf around the code and patiently waits for logs. I was shocked. How the hell is he so prolific?" https://frantic.im/mirage/

E se, anziché di "felicità" parlassimo più semplicemente di "gratificazione"? Cioè di una forma di contentezza derivata dalle esperienze sensoriali nella vita quotidiana, al contrario i sensazioni ben più sfuggenti e basate su un territorio impossibile da dominare? Money quote: "My hippy ex-boyfriend used to call that "taking a picture with your heart". We made fun of it then, but I find myself taking pictures with my heart all the time: some of it's trying to root myself in the present, but it's also often about trying to feel the world around me: the smell, the touch, the sounds, all the little stuff that makes a moment gratifying." https://www.patreon.com/CultureStudy/posts/how-to-fall-back-164373561