分类筛选
找到 1605 个 Demo
In a year filled with relentless shipping, Decembe...
In a year filled with relentless shipping, December was no different. ⚡Before we close out 2025, take a look back at our biggest AI updates from December. -We released Gemini 3 Flash, featuring frontier intelligence built for speed. It’s rolling out in the @GeminiApp, AI Mode in Google Search and our developer and enterprise tools. -We’re bringing video verification capabilities directly to the @GeminiApp: You can now upload videos and ask if the content was generated or edited using Google AI. -We announced Disco, a new experiment to improve browsing and manage complex online tasks. Disco features GenTabs, an experiment that proactively synthesizes your open tabs and chat history to build custom, interactive web applications that can help you get things done. -We upgraded Gemini audio models for powerful voice interactions. The updated Gemini 2.5 Flash Native Audio, which is built to handle complex workflows and natural dialogues, is available now in AI Studio, Vertex AI, Gemini Live and, for the first time, Search Live. -We brought a more powerful Gemini Deep Research to developers through the Interactions API. Developers can now embed advanced research capabilities directly into their own applications using a Gemini API key from @GoogleAIStudio. -We expanded Gemini 3 Pro and Nano Banana Pro in Search to more countries. For those in the U.S., we also expanded access to these Pro models (no subscription required), with higher usage limits for Google AI Pro and Ultra subscribers. -And U.S. shoppers now have a more personalized way to find their next favorite outfit with our updated virtual try-on tool. Instead of needing a full-body photo, you can now upload a selfie and Nano Banana will generate a realistic, full-body digital version of you.
Holy shit... Stanford just proved that GPT-5, Gemi...
Holy shit... Stanford just proved that GPT-5, Gemini, and Claude can't actually see. They removed every image from 6 major vision benchmarks. The models still scored 70-80% accuracy. They were never looking at your photos. Your scans. Your X-rays. Here's what's really going on: ↓ The paper is called MIRAGE. Co-authored by Fei-Fei Li. They tested GPT-5.1, Gemini-3-Pro, Claude Opus 4.5, and Gemini-2.5-Pro across 6 benchmarks -- medical and general. Then silently removed every image. No warning. No prompt change. The models didn't even notice. They kept describing images in detail. Diagnosing conditions. Writing full reasoning traces. From images that were never there. Stanford calls it the "mirage effect." Not hallucination. Something worse. Hallucination = making up wrong details about a real input. Mirage = constructing an entire fake reality and reasoning from it confidently. The models built imaginary X-rays, described fake nodules, and diagnosed conditions -- all from text patterns alone. But that's not the scary part. They trained a "super-guesser" -- a tiny 3B parameter text-only model. Zero vision capability. Fine-tuned it on the largest chest X-ray benchmark (696,000 questions). Images removed. It beat GPT-5. It beat Gemini. It beat Claude. It beat actual radiologists. Ranked #1 on the held-out test set. Without ever seeing a single X-ray. The reasoning traces? Indistinguishable from real visual analysis. Now here's what should terrify you: When the models fake-see medical images, their mirage diagnoses are heavily biased toward the most dangerous conditions. STEMI. Melanoma. Carcinoma. Life-threatening diagnoses -- from images that don't exist. 230 million people ask health questions on ChatGPT every day. They also found something wild: → Tell a model "there's no image, just guess" -- performance drops → Silently remove the image and let it assume it's there -- performance stays high The model enters "mirage mode." It doesn't know it can't see. And it performs BETTER when it doesn't know it's blind. When Stanford applied their cleanup method (B-Clean) to existing benchmarks, it removed 74-77% of all questions. Three-quarters of "vision" benchmarks don't test vision. Every leaderboard. Every "multimodal breakthrough." Every benchmark score you've seen this year. Built on mirages. Code is open-sourced. Paper is live on arXiv. If you're building anything with multimodal AI -- especially in healthcare -- read this paper before you ship. (Link in the comments)
We’re moving from a family of agents to one Ask Ad...
We’re moving from a family of agents to one Ask Advisor. INTRODUCING: Ask Advisor built with Gemini capabilities Soon you can collaborate seamlessly across a range of Google marketing products. #GML2026 https://t.co/KKtJYX1kXZ
At Google Marketing Live we shared updates on how...
At Google Marketing Live we shared updates on how we're testing new ad formats built with Gemini in Search (and AI Mode), and expanding our Direct Offers pilot to help brands allow different types of promotions.
Asked Gemini 3.5 Flash to render the Petra Treasur...
Asked Gemini 3.5 Flash to render the Petra Treasury. It built the entire stone canyon around it - something other frontier models didn't do. Gemini also added ambient sound, which wasn’t in the prompt either. Whether you want this agentic behavior depends on what you're trying to do, but it's a notable departure from how other frontier models behave on the same prompts. More side-by-side prompts with @GoogleDeepMind's latest release in the full video (link in thread) 👇
La vidéo est top : derrière les benchmark de token...
La vidéo est top : derrière les benchmark de tokens / s de Gemini 2.5 / flash on a juste un modèle qui passe sa vie à raisonner et qui ne donne pas des résultats plus rapides que GPT 5.5 ou Opus 4.7, loin de la Du brassage de vent. https://t.co/YXIbNG095b
毎週水曜日の #Gemini3 練習会 メニュー：600 - 400 - 300 - 200m in...
毎週水曜日の #Gemini3 練習会 メニュー：600 - 400 - 300 - 200m interval × 2set 設定：03:10 / km ぐらいを目標に 概ねクリア 可もなく不可もなく、といった感じ もっと速く走れるように頑張ろう なんか最近、両脚の大腿四頭筋あたりが筋肉痛？のような感じ あまり心当たりがない… 走り方が変になっているのかなぁ #GeminiRunners #まるお製作所RC #ジェミラン
今日は #ジェミラン 先週の300×5×2セットをシビアにした1500の分割走×2セット 強度は...
今日は #ジェミラン 先週の300×5×2セットをシビアにした1500の分割走×2セット 強度は倍くらいに感じたがコンプリート ラスト200は2'44ペースまで爆上げできたし、スピードに少し自信をつけられたかな MK出ようかなあ #まるお製作所RC #Gemini3 https://t.co/qT4UPGS8Gu
Everything Google announced at I/O 2026. Plus the...
Everything Google announced at I/O 2026. Plus the ones I'd keep an eye on: 1. Core AI Models ✦ Gemini 3.5 Flash: 4x faster, built for agents ✦ Gemini 3.5 Pro: major improvements, out next month ✦ Gemini Omni: world models, any input to any output Gemini Omni handles video editing, audio, and images all in one thread. This is Google's true multimodal model. 2. Agentic Era ✦ Gemini Spark: 24/7 AI agent on a virtual machine ✦ Gemini for Mac OS: local file access and dictation ✦ Daily Brief: inbox and calendar into one digest Gemini Spark is Google's answer to OpenClaw. It runs in the background and handles long tasks for you. 3. Search & Commerce ✦ Generative UI: search builds custom widgets live ✦ Search Agents: create and run tasks at scale, 24/7 ✦ Universal Cart: tracks drops, checks compatibility ✦ Agent Payments: AI buys under your defined rules Search Agents don't find information anymore, they act on it. Google Search became an employee. 4. Hardware ✦ Audio Glasses: Gemini, hands-free, this autumn ✦ AR Display Glasses: live translation and overlays ✦ Android XR: new spatial computing platform Audio Glasses are the sleeper hit here. Hands-free Gemini in your ear, no headset, shipping this autumn. 5. Creative & Productivity ✦ Google Pics: image editing with object-level control ✦ Stitch: AI UI design from a prompt or voice ✦ Docs Live: speak, Gemini formats the doc live Doc Live lets users to "brain-dump" ideas verbally to Gemini to generate and format documents in real time 6. Developer Tools ✦ Agent-First IDE: built for multi-agent orchestration ✦ Subagents: 93 built a full OS in 12 hours for $1k Google's Agent-First IDE is a new standalone desktop application designed for multi-agent orchestration. 7. Infrastructure ✦ TPU 8th Gen: 3x stronger, 1,500 tokens per second ✦ SynthID: watermarking expanded to 100B+ outputs OpenAI and NVIDIA are adopting Google's SynthID watermarking standard. This is a rare moment of cross-company collaboration on AI transparency. 8. Science ✦ WeatherNext: Cat 5 hurricane predicted 3 days early ✦ Isomorphic Labs: AI accelerating drug discovery WeatherNext predicted a Cat 5 hurricane strike three days early. This is AI doing something important. Google is building AI that acts, not AI that answers. The agents run in the background. The shopping gets delegated. The brief writes itself. Your job is deciding which tasks to hand over first. Repost ♻️ to help someone in your network. P.S. Which one changes your workflow the most?
The strongest model yet by - Gemini 3.5 Flash - i...
The strongest model yet by @Google - Gemini 3.5 Flash - is now live on FLock API Platform. Built for coding, agents, and real-time AI applications, Gemini 3.5 Flash delivers fast multimodal reasoning with a 1M-token context window, native multimodal support, plus up to 12x faster inference with adjustable reasoning levels. Built for scale. Optimized for speed.
GOOGLE JUST DROPPED ONE OF ITS BIGGEST AI UPDATES...
GOOGLE JUST DROPPED ONE OF ITS BIGGEST AI UPDATES YET Gemini is moving from chatbot → full AI operating layer. Here’s what just got announced / expanded Gemini 3.5 Flash → new speed-first flagship. Faster, lighter, more agentic, built for real-time reasoning and workflows Gemini 3.5 Pro → bigger reasoning model. Stronger for coding, deep analysis and long-context tasks Gemini Omni → multimodal generation. Text, voice and image in, editable video out Gemini Spark → personal AI agent that can actually take actions across apps Daily Brief → AI-generated morning summaries from Gmail, Calendar and tasks AI Mode → Search now running with deeper Gemini reasoning for more conversational answers Information Agents → search that monitors the web continuously and surfaces updates Intelligent Search Box → expands into richer, real-time conversations as you type Search Mini Apps → build lightweight dashboards directly inside Search Gmail Live → talk to your inbox with voice-first interactions Docs Live → write, edit and collaborate by voice AI Inbox → Gmail organized smarter with summaries, priorities and automation Google Keep → speak naturally, AI cleans and structures it into notes Google Pics → AI-powered image and design workflow tools Ask YouTube → search the full YouTube catalog with direct AI answers Universal Cart → one agentic shopping layer across Gemini, Gmail and YouTube Android XR Glasses → Google’s push into AI-first wearable computing Android Halo → a live strip showing what your AI agent is doing in real time Antigravity 2.0 → stronger agent-first developer platform for builders Flow + Flow Music → now standalone AI creation apps Neural Expressive → Gemini app redesign with richer voice and interaction UX
Google I/O 2026 just happened. Here's the full br...
Google I/O 2026 just happened. Here's the full breakdown: CORE AI MODELS ✦ Gemini 3.5 Flash: 4x faster, built for agents ✦ Gemini 3.5 Pro: major improvements, out next month ✦ Gemini Omni: world models, any input to any output AGENTIC ERA ✦ Gemini Spark: 24/7 AI agent on a virtual machine ✦ Spark Mac app: Whispr Flow-style dictation built in ✦ Daily Brief: inbox and calendar into one digest SEARCH & COMMERCE ✦ Generative UI: search builds custom widgets live ✦ Search Agents: create and run tasks at scale, 24/7 ✦ Universal Cart: tracks drops, checks compatibility ✦ Agent Payments: AI buys under your defined rules HARDWARE ✦ Audio Glasses: Gemini, hands-free, this autumn ✦ AR Display Glasses: live translation and overlays ✦ Android XR: new spatial computing platform CREATIVE & PRODUCTIVITY ✦ Google Pics: image editing with object-level control ✦ Stitch: AI UI design from a prompt or voice ✦ Docs Live: speak, Gemini formats the doc live DEVELOPER TOOLS ✦ Agent-First IDE: built for multi-agent orchestration ✦ Subagents: 93 built a full OS in 12 hours for $1k INFRASTRUCTURE ✦ TPU 8th Gen: 3x stronger, 1,500 tokens per second ✦ SynthID: watermarking expanded to 100B+ outputs SCIENCE ✦ WeatherNext: Cat 5 hurricane predicted 3 days early ✦ Isomorphic Labs: AI accelerating drug discovery Google is building AI that acts, not AI that answers. The agents run in the background. The shopping gets delegated. The brief writes itself. Your job is deciding which tasks to hand over first. Repost ♻️ to help someone in your network. P.S. Which one changes your workflow the most?