Ola todos! ecco le notizie dedicate al mondo IA del 07 Maggio 2026
Iniziamo con una scoperta che potrebbe far discutere: Google sembrerebbe aver installato silenziosamente un modello IA da 4 GIGABYTE su oltre un miliardo di computer tramite Chrome. Questo file verrebbe scaricato senza consenso, sollevando seri dubbi sulla privacy.
E mentre Google fa discutere, OpenAI sembrerebbe aver iniziato il rilascio di GPT Instant 5.5 per tutti gli utenti, promettendo risposte più brevi e ragionamenti migliorati. Ma la concorrenza non dorme: Kimi K2.6 sarebbe appena atterrato gratuitamente su Nvidia, capace di gestire 300 sub-agenti e 256mila token di contesto.
Passiamo ora al colosso del momento: Anthropic. L’azienda potrebbe aver appena spaventato Wall Street lanciando 10 AGENTI FINANZIARI capaci di sostituire gli analisti bancari, facendo crollare le azioni dei fornitori di dati. Per supportare questa espansione, avrebbero stretto un accordo con SpaceX per usare il supercomputer Colossus 1, raddoppiando i limiti di Claude. Inoltre, sembrerebbero aver lanciato la funzione ‘dreaming’ per i Managed Agents in preview. E se volete imparare a sfruttarli, avrebbero pubblicato un workshop gratuito per costruire aziende basate sugli agenti, oltre a rilasciare il loro primo esame ufficiale gratuito per Architetti IA. Infine, per chi usa Claude Code, il nuovo tool ‘WozCode’ potrebbe ottimizzare i flussi di lavoro riducendo gli sprechi.
Spostandoci sulle prestazioni locali, Qwen 3.6 sembrerebbe generare oltre 55 token al secondo su una semplice RTX 4060. Qualcuno lo avrebbe persino integrato nel browser Strawberry per un’esperienza totalmente privata e gratuita. A proposito di browser, un nuovo ‘Chrome MCP Server’ permetterebbe agli assistenti IA di controllare direttamente la vostra navigazione mantenendo login ed estensioni. Se invece preferite costruire da zero, sarebbe disponibile un nuovo tutorial per creare un assistente di ricerca basato su Grok.
Passiamo alla creatività. Kling 3.0 sembrerebbe pronto per il cinema, permettendo di sostituire green screen in 4K con pochi dollari. Per i creatori locali, un nuovo workflow ComfyUI permetterebbe di generare video con LTX2.3 usando solo 8GB di VRAM. Se vi serve il 3D, Runware avrebbe aggiunto 5 nuovi modelli text-to-3D alle sue API. Sul fronte audio, ‘StemDeck’ separerebbe le tracce dei video YouTube direttamente in locale senza registrazioni, mentre due nuovi paper su ‘Hibiki-zero’ prometterebbero traduzioni vocali simultanee con latenza bassissima.
Chiudiamo col mondo del lavoro. Stripe sembrerebbe aver creato una figura aziendale dedicata per affiancare i dipendenti e aiutarli ad automatizzare le proprie mansioni. Infine, per imparare nuovi software, ‘Spotit’ potrebbe guidarvi indicando esattamente dove cliccare direttamente dentro l’interfaccia dell’applicazione.
Salvate il video così da non perderlo, seguitemi sui miei canali social e sul nuovo podcast su spotify e mannaggia a voi se ancora non lo state facendo, condividete il video con amici e nemici e ola todos.
O Google acabou de transformar mais de 1 bilhão de computadores em depósito de IA.
— Felipe Demartini (@namcios) May 6, 2026
Inclusive o seu.
Sem pedir. Sem avisar. Sem um único popup.
O Chrome baixou 4GB de modelo de inteligência artificial no seu disco. O arquivo se chama weights.bin, são os pesos do Gemini Nano.… https://t.co/nY9BW7yKFV pic.twitter.com/aKR0eX4zPd
OpenAI is rolling out GPT Instant 5.5 to all ChatGPT users, with shorter responses, fewer hallucinations, stronger reasoning, and a clearer memory interface. pic.twitter.com/V3xQlF7OyD
— Byte (@ByteEcosystem) May 6, 2026
𝗞𝗶𝗺𝗶 𝗞𝟮.𝟲 𝗷𝘂𝘀𝘁 𝗱𝗿𝗼𝗽𝗽𝗲𝗱 𝗳𝗿𝗲𝗲 𝗼𝗻 𝗡𝘃𝗶𝗱𝗶𝗮 𝗮𝗻𝗱 𝗶𝘁 𝗰𝗼𝗱𝗲𝘀 𝗮𝗹𝗼𝗻𝗲 𝗳𝗼𝗿 𝟭𝟯 𝗵𝗼𝘂𝗿𝘀.
— Julian Goldie SEO (@JulianGoldieSEO) May 6, 2026
And it builds apps from a screenshot.
Here's what one model now does:
→ Holds 256k tokens for your whole codebase
→ Spins up 300 sub agents on big… pic.twitter.com/mJX1rvGEu8
Breaking: Anthropic just shipped 10 AI agents that do the work of investment banking analysts. The market priced in the disruption within hours.
— Aria Westcott (@AriaWestcott) May 6, 2026
FactSet fell 8.1%. Morningstar erased its gains and fell more than 3%. S&P Global and Moody's both saw sharp selling pressure.
The… pic.twitter.com/TIXHd5AGQO
SpaceX/xAI has just announced that it has signed an agreement with Anthropic to provide access to its Colossus 1 supercomputer.
— Sawyer Merritt (@SawyerMerritt) May 6, 2026
"Anthropic plans to use this additional compute to directly improve capacity for Claude Pro and Claude Max subscribers. As part of this agreement,… pic.twitter.com/kcOVCCWzRS
Live from Code with Claude: we're launching dreaming in Claude Managed Agents as a research preview.
— Claude (@claudeai) May 6, 2026
Outcomes, multiagent orchestration, and webhooks are now in public beta. pic.twitter.com/p4DFRzFEd8
Anthropic ha publicado un taller sobre cómo construir una empresa solo con Agentes IA.
— angel (@angeldot_) May 6, 2026
Agentes trabajando entre ellos, repartiéndose tareas y ejecutando procesos.
Gratis. Del equipo de Claude.
Lo he subtitulado al español.
Si quieres que la IA trabaje por ti, guarda esto 🔖 pic.twitter.com/3qBHdtpJuj
🚨 Anthropic just launched its first official AI certification And it's FREE !
— Dipanshu Kushwaha (@DipanshuKu55175) May 6, 2026
Here's everything you need to know 👇
📌 What it is:
The Claude Certified Architect, Foundations (CCA) launched on March 12, 2026
It's a proctored, 60-question exam testing real production… pic.twitter.com/DTRvc79pCn
Most discussions around Claude Code focus on model quality or capability.
— Parul Gautam (@Parul_Gautam7) May 6, 2026
But when you actually use it, the friction shows up earlier, more at the system level than the model itself.
Token limits get the blame, but a lot of it comes from inefficient sessions: repeated context,… pic.twitter.com/Tem7oTBQCJ
Qwen3.6 35B A3B model. 55+ tokens/sec. $300 GPU.
— AboveSpec (@above_spec) May 6, 2026
No, this isn't a server card. It's an RTX 4060 Ti 8GB.
Previously I posted that I 41 t/s on this gpu and that post blew up and went viral. I went back and made it 34% faster.
And now the speed doesn't drop with context depth at… pic.twitter.com/iMNi5mku62
🚨 Founders: free, private, 100% local AI is finally here
— Charles Maddock (@charles_maddock) May 6, 2026
I downloaded Qwen 3.6 and ran it on my MacBook inside Strawberry browser.
I couldn't believe it.
We now have 100% free and private AI that is on par with the best models from OpenAI, Anthropic and Google. (see video!)… pic.twitter.com/DQH5AZxqW4
又挖到一个很能打的 MCP 服务器:Chrome MCP Server。它基于 Chrome 插件,让 AI 助手直接接管你的浏览器。
— Joruno (@wsl8297) May 6, 2026
不仅能做复杂的浏览器自动化、内容分析、语义搜索,还能原样保留你的登录状态、扩展和各类浏览器设置。
GitHub:https://t.co/lBnnno7Z55
主要亮点:
– 直接用你正在用的… pic.twitter.com/wfStJ3UG5L
Lets Build a Grok-Powered agentic research assistant with LangGraph — here's the architecture that makes it work.
— Marktechpost AI Dev News ⚡ (@Marktechpost) May 6, 2026
Most agent demos stop at a single LLM call. This one goes deeper:
< Inference: Groq's OpenAI-compatible endpoint with `llama-3.3-70b-versatile` — just swap the… pic.twitter.com/8ozAfWbPky
AI vfx is now ready for big screen
— el.cine (@EHuanglu) May 6, 2026
you can now replace green screen and props in 4K quality on Kling 3.0.. big team's months of work now can be done on a laptop in mins
$100,000 4K effects now can be done for $5
here are some 4K raw clips with prompts: pic.twitter.com/0PaWvwOwv4
VRAM 8GB環境でLTX2.3の動画生成を実現するComfyUIワークフロー。低解像度の基本生成と別工程でのアップスケール・補間によりVRAM負荷を低減。テキスト・画像からの生成に対応。安定動作には32GB以上のシステムメモリを推奨。#StableDiffusion #ComfyUI
— R@aiaicreate (@aiaicreate) May 6, 2026
URLはリプ⬇️ pic.twitter.com/y2CHhgJicH
5 text-to-3D models just dropped on Runware 🧊
— Runware (@runware) May 6, 2026
– Meshy 6, Tripo H3.1, Hyper3D Rodin Gen-2, Hunyuan 3D 3.1 Pro & Rapid
– text into production-ready 3D assets, no 3D apps required
– available now via our API.
here’s what they do 👇
which one are you trying first? pic.twitter.com/rOMRHZ75Pi
粘贴 YouTube 链接,把歌拆成人声、鼓、贝斯、吉他、钢琴和其他六轨,全程在自己机器上跑,不用注册也不用上传。https://t.co/3N0LDca5st
— Geek Lite (@QingQ77) May 6, 2026
StemDeck 用 Demucs 模型把 YouTube 音频拆成六轨,给了一个类似 DAW… pic.twitter.com/WLXDKPo80O
Two speech-to-speech papers from our team accepted at ICML, including one spotlight: Hibiki-zero and Moshi-RAG.
— Neil Zeghidour (@neilzegh) May 6, 2026
First, Hibiki-zero: speech-to-speech translation that jointly optimizes translation quality and latency with RL. https://t.co/c4Yt4f201T
Breaking 🤯
— Swadesh Kumar (@swadeshkumar_) May 6, 2026
Stripe just created a role that didn't exist 12 months ago (and they're paying multiple six figures for it)
It's called the Forward Deployed AI Accelerator.
They are hiring AI-native individuals to work directly with their marketing teams to fundamentally change… pic.twitter.com/DljEtrVeMp
ChatGPT accidentally created a huge problem.
— Kshitij Mishra | AI & Tech (@DAIEvolutionHub) May 6, 2026
People stopped learning software.
Now they just copy prompts…
watch tutorials…
and pray it works.
This startup may have just killed that behavior.
It’s called Spotit.
Found it here:https://t.co/rfbLEMMUxL
And it’s one of the… pic.twitter.com/jB7lBNetXT
