Francesco Marradi

A life on wings... and a pair of foreplanes and afterburners ;) Città di Castello

[CentroStudiAmericani @centrostudiusa]

70 miliardi di $: è il budget USA per la Space Force nel 2027, quasi il doppio di tutta la difesa italiana. Lo spazio è il nuovo fronte strategico globale.

Leggi il paper https://t.co/vlKcG1FuMP

[PHALANX @PHALANX_LTD]

Race to Rabaul is our logistics-focused game on the Pacific Theatre of WW2, made by legendary designer Volko Ruhnke.

You will need to manage supply lines over land and sea, shifting transports, men, food and ammo to where they are needed.

The beautiful and visually clear point-to-point map shows your pathways to victory - one island at a time.

Check out the game here:
https://t.co/aJXFvaFaAC

[Pier Luigi Gentili @Pier_Complexity]

Al @SalonedelLibro c'era anche il mio ultimo #libro: "IL MONDO È COMPLESSO: PARLIAMONE!" edito da Morlacchi.
#divulgazione della #scienza
@UniperugiaNews

[Enrico Pagliarini @e_pagliarini]

L'ospite di questa puntata è @AlfonsoFuggetta Buon ascolto! https://t.co/kH1smVmqGg

[Wargaming Weekly @wargamingweekly]

"At the 2026 Games of War conference, one point was clear: the pace of modern warfare, especially in Ukraine, is compressing the timeline between observing conflict, analyzing it, and acting on it. The traditional model–study first, apply later–is no longer sufficient."

[Thinking Machines @thinkymachines]

People talk, listen, watch, think, and collaborate at the same time, in real time. We've designed an AI that works with people the same way.

We share our approach, early results, and a quick look at our model in action.

https://t.co/AFJZ5kH7Ku

[Luiza Jarovsky, PhD @LuizaJarovsky]

🚨 The recently launched book "Legal, Ethical, and Technical Dilemmas in Military AI" deals with one of the most challenging topics in AI governance today.

Download it for free below.

[Patrycia Centeno @PoliticayModa]

Homenaje de Carlos III al divulgador científico y naturista David Attenborough.

En Buckingham no hacen vídeos, hacen CINE!!!! 😍😍😍😍

[Paroles d'auteurs @Paroles_auteurs]

Merci à Christophe Clavé pour cet éclairage sur l’appauvrissement de la langue et la ruine de la pensée 🙌

"La disparition progressive des temps (subjonctif, passé simple, imparfait, formes composées du futur, participe passé…) donne lieu à une pensée au présent, limitée à l’instant, incapable de projections dans le temps.
La généralisation du tutoiement, la disparition des majuscules et de la ponctuation sont autant de coups mortels portés à la subtilité de l’expression.
Supprimer le mot «mademoiselle» est non seulement renoncer à l’esthétique d’un mot, mais également promouvoir l’idée qu’entre une petite fille et une femme il n’y a rien.
Moins de mots et moins de verbes conjugués c’est moins de capacités à exprimer les émotions et moins de possibilité d’élaborer une pensée.
Des études ont montré qu’une partie de la violence dans la sphère publique et privée provient directement de l’incapacité à mettre des mots sur les émotions.
Sans mot pour construire un raisonnement, la pensée complexe chère à Edgar Morin est entravée, rendue impossible.
Plus le langage est pauvre, moins la pensée existe.
L’histoire est riche d’exemples et les écrits sont nombreux de Georges Orwell dans « 1984 » à Ray Bradbury dans « Fahrenheit 451 » qui ont relaté comment les dictatures de toutes obédiences entravaient la pensée en réduisant et tordant le nombre et le sens des mots.
Il n’y a pas de pensée critique sans pensée. Et il n’y a pas de pensée sans mots.
Comment construire une pensée hypothético-déductive sans maîtrise du conditionnel ? Comment envisager l’avenir sans conjugaison au futur ? Comment appréhender une temporalité, une succession d’éléments dans le temps, qu’ils soient passés ou à venir, ainsi que leur durée relative, sans une langue qui fait la différence entre ce qui aurait pu être, ce qui a été, ce qui est, ce qui pourrait advenir, et ce qui sera après que ce qui pourrait advenir soit advenu ? Si un cri de ralliement devait se faire entendre aujourd’hui, ce serait celui, adressé aux parents et aux enseignants : faites parler, lire et écrire vos enfants, vos élèves, vos étudiants.
Enseignez et pratiquez la langue dans ses formes les plus variées, même si elle semble compliquée, surtout si elle est compliquée. Parce que dans cet effort se trouve la liberté. Ceux qui expliquent à longueur de temps qu’il faut simplifier l’orthographe, purger la langue de ses «défauts», abolir les genres, les temps, les nuances, tout ce qui crée de la complexité sont les fossoyeurs de l’esprit humain. Il n’est pas de liberté sans exigences. Il n’est pas de beauté sans la pensée de la beauté."
Christophe Clavé

[Simon's Cat 🐾 @SimonsCat]

Sir David Attenborough turns 100 today 🎉💯 A champion for planet earth, guiding us through jungles, oceans and other wonders of our planet. 🌍 #HappyBirthdayDavidAttenborough

[Luca @Digione_79]

Su Gilles si è detto, si dice e si dirà sempre di tutto e di più. Per me è stato e resterà quella febbre che non passa e non passerà mai.
Salut aviatore. 🇨🇦 🙏

[Francesco Marradi @FMarradi]

@sabbatini Ciao Aviatore! Mai dimenticato!

[Alberto Sabbatini @sabbatini]

Oggi 8 maggio un pensiero
per Giles nel ricordo della sua scomparsa a Zolder.
Voglio ricordarlo con questa foto in azione di Ercole Colombo che riassume perfettamente i’essenza del suo modo di guidare così spettacolare, che è quello che ci ha sempre emozionato.
#Villeneuve

[Phil Venables @philvenables]

Everything in security comes down to economics eventually.

https://t.co/SIdbfPHYwZ

[MIT.nano @mit_nano]

A new study from @MIT researchers could help enable next-generation lidar sensors that are compact, durable, and have no moving parts. The key advance is a novel design for a silicon-photonics chip, which is a semiconductor device that manipulates light. https://t.co/uCo6quEwxF

[CSIS @CSIS]

A new report from @csisfutures introduces a spatial-temporal approach to track and measure a growing and sustained campaign by China to apply pressure on Taiwan through both its air and maritime forces.

Read the report here: https://t.co/KojEEt0gmM

[Cambridge University Press - History @cambUP_History]

New Cambridge Element, Using Generative AI in Historical Practice, by Yaniv Fox, out now! Read for free for the next 2 weeks at
https://t.co/5KMfKFxuYZ
#cambridgeelements #history

[Gerardo L. Munck @GerardoMunck]

Political Scientists on AI

This is an interesting collection of essays prepared by @persily and @j_a_tucker that discusses how AI is reshaping politics & how political scientists study politics. A timely contribution to a key discussion.

Open Access: https://t.co/UvubILo4vZ

[Akshay 🚀 @akshay_pachaar]

Google DeepMind dropped a paper that should scare every agent builder.

It's the first systematic framework for a threat that barely existed two years ago: adversarial content engineered to hijack AI agents browsing the web.

They call them AI Agent Traps. The paper maps six distinct attack surfaces.

1) Content Injection Traps (perception)

Invisible CSS, hidden HTML, steganographic payloads inside images. The agent parses it, humans never see it. One study showed simple HTML injections hijack web agents in up to 86% of scenarios.

2) Semantic Manipulation Traps (reasoning)

No overt commands. Just biased phrasing, framing, and contextual priming that skew the agent's synthesis. LLMs inherit human cognitive biases, and attackers can weaponize every one of them.

3) Cognitive State Traps (memory and learning)

Poison the RAG corpus. Corrupt long-term memory. One study achieved over 80% attack success with less than 0.1% poisoned data.

4) Behavioural Control Traps (action)

Jailbreaks embedded in external resources. Data exfiltration prompts hidden in emails. Sub-agent spawning that tricks an orchestrator into instantiating attacker-controlled agents inside the trusted control flow.

5) Systemic Traps (multi-agent dynamics)

This is where it gets scary. A single fake news headline could trigger a synchronized sell-off. A compositional fragment trap splits a payload across sources, so each fragment looks benign until agents aggregate them.

6) Human-in-the-Loop Traps

The agent becomes the vector. The target is you. Invisible prompt injections have already caused summarization tools to faithfully repeat ransomware commands as "fix" instructions.

The core insight is uncomfortable.

By altering the environment instead of the model, attackers weaponize the agent's own capabilities against it. Training-time defenses cannot solve an inference-time problem.

The paper closes by calling for automated red-teaming that can probe these vulnerabilities at scale. That same shift is already happening on the offense side.

Strix is an open-source project doing exactly this for web apps. AI agents that act like real hackers, running your code dynamically, finding vulnerabilities, and validating them with actual proof-of-concepts.

24k stars on GitHub. Apache 2.0 licensed.

The agents writing your code need to be tested by agents trying to break it.

I've shared the link to the paper and Strix GitHub repo in the replies

[Andrea Bernardi @Bernardi_UK]

Siamo molto lieti di annunciare che il CASD Scuola Superiore Universitaria ha ottenuto due finanziamenti EDF 2025 (European Defense Fund) di durata triennale. Il corpo docente e ricercatore sarà impegnato in due consorzi che hanno ottenuto dalla Commissione Europea 74ml di euro.