Free email without sacrificing your privacy

Gmail is free, but you pay with your data. Proton Mail is different.

We don’t scan your messages. We don’t sell your behavior. We don’t follow you across the internet.

Proton Mail gives you full-featured, private email without surveillance or creepy profiling. It’s email that respects your time, your attention, and your boundaries.

Email doesn’t have to cost your privacy.

Ditch the Gmail data grab

Today:

• GPT-5.1 Launches with New Instant & Thinking Modes

• Anthropic Unveils $50B Plan to Build U.S. AI Data Centers

• DeepMind Trains Vision AI to “See” More Like Humans

• Microsoft Builds Twin AI Supercenters in Wisconsin and Atlanta

• Meta Builds Eco-Friendly AI Data Center in Wisconsin

GPT-5.1 Launches with New Instant & Thinking Modes

GPT-5.1 upgrades GPT-5 with warmer, clearer conversation, smarter instruction-following, and adaptive reasoning. It introduces Instant and Thinking variants, improves speed on easy tasks, depth on hard ones, and adds stronger personalization controls so users can easily tune ChatGPT’s tone, style, and behavior across all chats.

KEY POINTS

• Two upgraded models: GPT-5.1 Instant is more playful and fast, while GPT-5.1 Thinking spends extra time on hard problems and explains ideas more clearly.

• Adaptive reasoning helps the system “think longer” only when needed, boosting math and coding accuracy without slowing simple replies.

• New tone controls and personality presets (like Professional, Friendly, Quirky) plus finer sliders for warmth, brevity, and emoji use make ChatGPT easier to shape to each user.

Why it matters

These changes make ChatGPT feel more human and more helpful at the same time. You get better answers on tough tasks, quicker replies on easy ones, and a voice that can match your needs whether you want concise work help, a friendly chat, or a playful creative partner, all without constant tweaking.

Anthropic Unveils $50B Plan to Build U.S. AI Data Centers

Anthropic plans to invest $50bn in US computing infrastructure, building datacenters in Texas and New York with Fluidstack. The Claude maker says the expansion will supply the computing power needed for more advanced AI systems that can speed up research and help solve difficult problems.

KEY POINTS

• Anthropic will spend $50bn on new US datacenters, mainly in Texas and New York.

• The company is partnering with London-based Fluidstack to design and build these huge computing sites.

• The extra computing power is meant to support more powerful AI that can aid scientific discovery and complex problem-solving.

Why it matters

This is a very large bet on the future of AI. New datacenters mean more “computer factories” that run models like Claude. They can boost research, create local jobs, and raise questions about power use, energy sources, and how AI growth is managed in the US.

DeepMind Trains Vision AI to “See” More Like Humans

Google DeepMind shows that changing how vision AI groups images can make it think more like people. By training models on human “odd one out” choices, they reshape maps of concepts, boosting agreement with human judgments, learning from few examples, and reliability under changing conditions.

KEY POINTS

• Researchers compare humans and vision AI on “odd one out” image tests and find many cases where models group things in ways that feel wrong to people.

• They train a small “teacher” model on human choices, use it to create a huge fake-but-human-like dataset, then retrain larger “student” models to organize concepts more like we do.

• The aligned models agree with people more often, handle new image types better, and can learn new categories from very few images, while also showing human-like uncertainty.

Why it matters

This work makes AI that “sees” closer to how we see. When models group objects in human-like ways, their decisions become easier to predict and trust. That means fewer strange mistakes, better performance in real-world settings, and a clearer path to vision systems that match human common sense instead of fighting it.

🧠RESEARCH

Grounding Computer Use Agents on Human Demonstrations

Researchers built GroundCUA, a huge dataset of screenshots with human-labeled buttons and menus from many desktop apps. It trains GroundNext models to follow written instructions by picking the right on-screen items. With careful data and some reward-based training, these models match or beat rivals while using far less data overall.

Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5B

VibeThinker-1.5B is a small AI model trained with a two-step process: first explore many different answers, then strongly reward the correct ones. Despite its size, it beats or matches much larger systems on math and coding tests in real benchmark competitions, showing cheap, compact models can still reason very well.

IterResearch: Rethinking Long-Horizon Agents via Markovian State Reconstruction

IterResearch is a research agent design that tackles long, complex tasks by regularly cleaning and rewriting its notes instead of stuffing everything into one giant context. It is trained to explore efficiently, which lets it scale to thousands of steps and greatly outperform earlier open-source research systems on tough benchmarks.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

AI Avatar ML- Create AI Photos Of Yourself - a web-based studio for making AI photos of yourself

AI Beauty Analyzer – Beauty Score Calculator - web tool that rates facial attractiveness from an uploaded photo

AI Cheat Check – The Checker AI - web-based demo built for educators to flag likely AI‑generated text in student work

📲SOCIAL MEDIA

🗞️MORE NEWS

• Microsoft is building linked AI datacenters in Wisconsin and Atlanta that share data at near-light speed, pack GPUs (special AI chips), use low-water cooling, and act together like one giant computer to train powerful models.

• Meta is building an artificial intelligence (AI) data center in Beaver Dam, Wisconsin, creating jobs, helping local families, using cooling that needs almost no water, and restoring wetlands and prairie to protect water and animals.

• IBM now builds quantum computer chips in a leading New York chip factory, using large silicon disks, round-the-clock processing, and close teamwork to speed design cycles, improve chip quality, and scale toward practical quantum machines.

• ElevenLabs launched the Iconic Marketplace, a platform where companies can license famous voices like Sir Michael Caine’s. It aims to protect performers’ rights, use consent, and help brands create new stories with trusted, recognizable voices.

• Marble is a new AI tool that lets anyone turn text, images, video, or rough 3D layouts into detailed 3D worlds, then edit, grow, combine, and export them for games, films, design, or robotics simulations.