Clear communicators aren't lucky. They have a system.

Here's an uncomfortable truth: your readers give you about 26 seconds.

Smart Brevity is the methodology born in the Axios newsroom — rooted in deep respect for people's time and attention. It works just as well for internal comms, executive updates, and change management as it does for news.

We've bundled six free resources — checklists, workbooks, and more — so you can start applying it immediately.

The goal isn't shorter. It's clearer. And clearer gets results.

Get the free toolkit

Hey everyone,

I’ve been glued to the updates coming out of CES 2026, and I wanted to break down the three biggest stories that really caught my eye.

There’s a lot of noise out there, but here are the things that I think will actually change how we live and work.

Today:

• Why AGI Is a Fantasy

• Samsung Puts Google AI on 800M Devices

• Nvidia Gives Robots a Brain Upgrade

• Alexa Comes to Your Web Browser

• Top Researcher Leaves OpenAI

• Millions Turn to ChatGPT for Medical Advice

Lee Cronin "Sam Altman Is Delusional, Hinton Needs Therapy, P(Doom) Is Nonsense"

Chemist Lee Cronin argues that AI is nowhere near life, not in agency, not in creativity, not in consciousness. True intelligence, he says, evolves from life’s struggle to persist, driven by causation and selection long before biology existed. Unlike AI, life creates complexity at scale through recursive processes, memory, and imagination. 

Cronin proposes that intelligence emerges from this deep causal layering not from matrix math. He debunks AI doomsday fears, saying superintelligent AGI is a fantasy with no mechanism. For him, AI is a brilliant tool, not a mind. Life is the real miracle, and chemistry is the engine behind it.

WATCH THE VIDEO ON YOUTUBE

NVIDIA is trying to make “robotics” feel like “apps”

NVIDIA’s big move isn’t just “new robots”, it’s the attempt to standardize the robot-building stack the same way mobile standardized around app stores + SDKs.

They announced new open “physical AI” models (their term for models that can perceive the real world, reason about it, and plan actions), plus tooling to make training/evaluating robots less of a bespoke science project. Jensen Huang even framed it as a “ChatGPT moment for robotics,” which is a bold line. But I get what he’s pointing at: the tools are becoming packaged, shareable, and reproducible instead of handcrafted.

What dropped:

• Cosmos models (Transfer 2.5 + Predict 2.5) for world simulation / synthetic data and testing robot policies in simulation, plus Cosmos Reason 2 (a vision-language reasoning model aimed at “see → understand → act”).

• Isaac GR00T N1.6, a vision-language-action model aimed at humanoids (full-body control + stronger contextual reasoning).

• Isaac Lab-Arena, an open-source evaluation / benchmarking framework that plugs into known benchmarks (so robot skills get tested more like software).

• OSMO, an orchestration layer to run robotics workflows across mixed compute (workstations + cloud), and they say it’s already integrated into Microsoft’s Azure Robotics Accelerator toolchain.

• A practical edge angle too: Jetson T4000 (Blackwell-based) positioned as a higher-efficiency module for robots/autonomous machines.

The part that matters: they’re pushing hard on “open models + shared evals + repeatable pipelines.” If that actually sticks, it’s how you go from “cool demo robot” to “robotics developer ecosystem.” 

Alexa+ is now in the browser and Amazon wants it to do things, not just answer

Amazon is rolling out Alexa.com for Alexa+ Early Access users, which effectively turns Alexa into a web-based AI assistant (voice and keyboard-first). This is Amazon admitting what we all learned in 2024–2025: assistants can’t live in one place. They have to follow you across devices.

A couple details worth caring about:

• Amazon frames Alexa+ as combining “info with real-world actions” (to-dos, calendar updates, smart home control, reservations, etc.).

• They also emphasize continuity — your context/prefs carry across interfaces (voice ↔ mobile ↔ web).

• The web experience includes uploading things like documents/images/emails so Alexa can extract useful info (e.g., turn a recipe into a shopping list or sync events to calendars), plus deeper shopping + smart home hooks.

My read: this is less about “Alexa becomes ChatGPT,” and more about “Amazon finally ships the assistant UI people actually use at work: a chat box on a laptop.” And if they can reliably execute tasks (not just talk), it becomes sticky in a way pure chatbots often aren’t. 

Samsung says it’ll ship 800 million “Galaxy AI” devices in 2026 (powered largely by Gemini)

This is the most “quietly enormous” headline. Samsung plans to double the number of mobile devices with Galaxy AI features to 800 million units in 2026 (up from 400 million), with many of those features powered by Google’s Gemini, alongside Samsung’s own Bixby for some tasks. 

Two things jumped out:

• Samsung’s co-CEO T.M. Roh said they want AI applied across all products, functions, and services “as quickly as possible.”

• On actual usage, he said search is the most-used AI phone feature, with editing/productivity tools plus translation and summarization also common, basically: less “AI magic,” more “AI utility.”

If Samsung pulls off 800M units, that’s not just a Samsung story. It’s a distribution story for Gemini, because it normalizes a default “AI layer” for everyday phone behavior at massive scale. 

🧠RESEARCH

NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos

Researchers created NeoVerse, an AI that builds detailed, moving 3D worlds using standard videos from a single camera. By avoiding the need for complex, multi-angle data, it makes creating realistic 4D environments cheaper and faster. This breakthrough allows easier generation of 3D scenes from everyday "in-the-wild" footage.

Youtu-Agent: Scaling Agent Productivity with Automated Generation and Hybrid Policy Optimization

Youtu-Agent is a new system that automatically builds and improves AI assistants, removing the need for tedious manual setup. It allows AI to create its own tools and learn from experience using a mix of practice and reinforcement learning. This makes deploying smart, adaptable agents much faster and more efficient.

Fast-weight Product Key Memory

Current AI struggles to remember long conversations without getting slow. This new "Fast-weight" memory updates itself instantly while the AI reads, acting like a dynamic scratchpad. This allows models to handle massive amounts of text—up to 128,000 words—even if they were only trained on short snippets, significantly boosting performance.

🛠️TOP TOOLS

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

Blend AI – AI Photo Editing Suite - web and mobile AI photo editor built for sellers and creators.

BlipCut – AI Video Translator Tool - browser‑based AI tool for localizing videos: it transcribes, translates, clones voices, and lip‑syncs speech so the on‑screen speaker appears to talk naturally in the target language.

BlueWillow AI – AI Image Generator - text‑to‑image generator that turns short prompts into logos, graphics, and photo‑realistic visuals.

📲SOCIAL MEDIA

🗞️MORE NEWS

OpenAI Loses Top Researcher Jerry Tworek, a key researcher behind GPT-4 and ChatGPT, has left OpenAI after seven years. He reportedly left due to disagreements with CEO Sam Altman’s focus on pushing products over pure research. Tworek was instrumental in building the company's new reasoning models, and his exit signals growing tension within the AI giant.

ChatGPT’s New Role: Doctor? A surprising 5% of all ChatGPT messages are now health-related, with 40 million Americans using it daily for medical advice. This surge is largely driven by people trying to understand insurance bills or check symptoms after losing health subsidies. Experts warn that relying on the chatbot is risky because it can still make confident but dangerous mistakes.

NVIDIA’s New Supercomputer NVIDIA has unveiled "Rubin," a new computing platform designed to power the next generation of AI. It promises to be much more efficient, cutting the cost of running AI models by ten times compared to current tech. This massive upgrade aims to help companies build smarter AI systems without needing as many expensive chips.

AI That Drives Like a Human NVIDIA also launched "Alpamayo," a family of free tools to help self-driving cars think before they act. Instead of just reacting to the road, these systems use "reasoning" to handle complex situations more like a human driver would. This software acts as a teacher, helping train autonomous vehicles to be safer and smarter.

A Small But Mighty AI Model A research institute in Abu Dhabi released a new AI model called Falcon H1R that punches way above its weight. Despite being much smaller than its competitors, it claims to match the performance of models seven times its size. This efficiency comes from a unique design that allows it to process data faster than standard AI models.

Microsoft Faces "Microslop" Backlash Social media users are mocking Microsoft with the nickname "Microslop" after the CEO wrote a blog post praising AI that many found tone-deaf. The term "slop" refers to low-quality, AI-generated content that is flooding the internet and annoying users. This trend highlights a growing frustration with tech companies forcing AI features into products where people don't want them.