Good morning! Here's what's happening in AI today:
OpenAI rolls out GPT-Live, a new generation of voice models inside ChatGPT
xAI releases Grok 4.5, its first model built specifically for coding and agents
Google launches Video Remix to transform your videos with AI
How to set up Hermes Agent the right way
And 4 new AI tools worth trying today
AI TOOL
OpenAI introduced GPT-Live, a new generation of voice models for natural human-AI interaction, rolling out in ChatGPT now.
GPT-Live is a new class of voice models designed specifically for natural conversation, not just text-to-speech layered on top of a language model.
The rollout starts in ChatGPT, bringing a more human-feeling voice interaction to the app most people already use daily.
The models are built for natural human-AI interaction, meaning they handle conversation flow, interruptions, and back-and-forth more fluidly than previous voice features.
OpenAI describes this as a new generation of voice, signaling a meaningful step up from the Advanced Voice Mode that shipped previously.
Voice has always been the interface most people find natural, and GPT-Live is OpenAI's most serious attempt at making AI conversation feel genuinely human. If you have tried voice mode before and found it robotic or slow, this is worth revisiting. It is rolling out in ChatGPT now and available to try immediately.
AI MODEL
xAI released Grok 4.5, its first model trained specifically for coding and agents, built with Cursor and designed to deliver frontier intelligence at leading speeds.
Grok 4.5 is the first xAI model purpose-built for coding and agent tasks, trained in collaboration with Cursor for deep code understanding.
The model targets frontier intelligence at competitive speeds and cost efficiency, making it practical for developers who run high-volume agent workflows.
Being trained with Cursor means Grok 4.5 understands how developers actually work inside modern coding environments, not just isolated code snippets.
Available now via the xAI API, Grok 4.5 gives developers and agent builders direct access to xAI's most code-focused model to date.
Coding models are becoming the battleground for AI labs. Grok 4.5 is xAI's answer — purpose-built for the task rather than adapted from a general model. For developers building agents or working with Cursor, this is a model worth testing against whatever you are running now.
AI VIDEO
Google introduced Video Remix in Google Photos, a Gemini Omni-powered tool that lets you reimagine your videos in new styles using a library of easy-to-use templates.
Open Google Photos, go to the Create tab, and select Video Remix to start transforming any video in your library without any editing experience required.
Templates let you apply cinematic lighting, swap backgrounds, add stylized effects like watercolor overlays, and more in just a few taps.
The feature is powered by Gemini Omni, meaning the style transformations are AI-generated and adapt to the content of your specific video.
Video Remix lives inside Google Photos, an app already on most Android phones, so there is no new download or signup required to access it.
Video editing has always required either skill or expensive software. Video Remix puts creative transformation into an app most people already have on their phone. You do not need to know how to edit - you pick a style, and Gemini handles the rest. For anyone who shoots videos on their phone and wants them to look more polished, this is one of the most accessible AI video tools released this year..
HOW TO AI
I've watched more people install an AI agent and never open it again than actually build it into their day. They run the install command, ask it one question, and go back to whatever they were using before.
That's not a tool problem. That's a setup problem.
The people who actually get value from Hermes aren't the ones who installed it fastest. They're the ones who gave it a model, a place to live outside the terminal, and a reason to remember things. Installing an agent takes a minute. Actually using one takes five small decisions.
Here's the setup, stripped down to what actually matters.
Run one command, not ten decisions
Skip overthinking the install. Download the desktop app for macOS or Windows, or run one line in your terminal:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash. That's the whole step.
Pick a model, not "the best" model
There isn't one correct answer here, and looking for it wastes time.
Type hermes model and choose from 30+ providers including Claude, GPT, Gemini, DeepSeek, and local models. No lock-in.
If the first one feels wrong, switch it tomorrow.
Talk to it before you configure anything else
Don't jump to gateways and skills before the core loop even works.
Type hermes in your terminal and start talking.
Let it read your files, run commands, and remember the conversation. Get that one loop working before you add anything on top.
Give it somewhere to live besides your terminal
An agent stuck in a terminal window is an agent you'll forget about by Friday.
Run hermes gateway setup and connect it to Telegram, Discord, Slack, WhatsApp, Signal, or Email. Now it lives where you already are, not in a window you have to remember to open.
Let it build its own shortcuts
This is the part that makes it worth keeping. Run hermes skills browse to pull from 200+ reusable workflows, or let Hermes create its own from how you actually use it. This is also what gets it faster over time.
It isn't a static script. It's building a memory of how you work.
Know these before something breaks
Run hermes doctor when something goes wrong. Set your terminal to Docker for safer sandboxing.
Type /voice on for hands-free mode.
Use hermes continue to pick up exactly where you left off. Small commands, but they're the difference between fixing something in ten seconds and starting over.
None of this requires you to become a developer. Hermes runs under an MIT license, so you own your setup outright, and nothing leaves your machine unless you connect it somewhere yourself. The actual shift isn't installing the agent. It's giving it enough structure to become something you actually return to.
Reactor launched LingBot World 2, a real-time world model with richer interactivity, sharper visuals, and advanced camera controls, available now via API on Reactor.
BytePlus released Dola Seedream 5.0 Pro via API, an image generation model built for precision editing, multilingual creation, and production-ready visual assets.
Nous Research moved Hermes Agent to the cloud, letting you spin up a fully configured AI agent in two clicks and 60 seconds with team access controls built in.
⚡ Grok: xAI's first model built specifically for coding and agent tasks.
🎙️ ChatGPT: Try GPT-Live, OpenAI's new generation of voice models.
🎬 Google Photos: Transform your videos with AI-powered styles and templates.
🤝 Tavus: Deploy an AI agent as a real participant in your Google Meet calls.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial