Good morning! Here's what's happening in AI today:
Meta releases Muse Glimmer, an open-weight 30B model built for local agents
xAI's Grok rolls out Voice Connectors free for everyone, no setup required
OpenAI expands Daybreak with GPT-5.6-Cyber for authorized cybersecurity work
How to let Claude publish content for you, right from your phone
4 new AI tools worth trying today
AI MODEL
Meta unveiled Muse Glimmer, an open-weight 30B-parameter model optimized for local, always-on agent workflows, releasing the model weights under a permissive Apache 2.0 license and designing it to run entirely on consumer hardware like a Mac or a PC with a performant GPU.
Muse Glimmer delivers strong performance on key agentic use cases and benchmarks compared with leading models in its size category, aiming to compete with much larger closed models.
The model runs entirely on consumer hardware, meaning developers can deploy always-on local agents without cloud inference costs or sending data off-device.
Releasing under Apache 2.0 keeps the license fully permissive, allowing commercial use, modification, and redistribution without the restrictions some open-weight releases carry.
A capable 30B model that runs locally on a Mac or gaming PC changes the calculus for anyone building always-on agents, since it removes both the cloud cost and the privacy tradeoff of sending every request to an API. The Apache 2.0 license means startups and hobbyists alike can build on it commercially without legal friction. For a market dominated by proprietary frontier models, an open, locally-runnable option this capable is a meaningful shift in what's accessible outside big labs.
AI TOOL
xAI rolled out Voice Connectors in Grok, letting anyone generate voice memos, turn daily events into a personalized podcast, or set up automations for a daily brief, now live for everyone on iOS, Android, and Web with no setup required.
Voice Connectors can turn a day's worth of events, notes, or updates into a personalized podcast-style summary, giving users an audio format instead of scrolling through text.
The feature also supports building automations for a recurring daily brief, so a summary can generate and deliver on its own schedule without a manual prompt each time.
Availability spans iOS, Android, and Web simultaneously with zero configuration, a broader and faster rollout than most feature launches that stagger by platform.
Turning AI output into audio format taps into a genuinely underused mode for AI assistants, most of which still default to text-first interactions. A zero-setup, all-platform launch means adoption friction is close to nothing, which matters more for a feature like this than raw capability. For anyone who consumes information via podcasts or audio during commutes or workouts, this turns Grok into a passive information source instead of something that requires active screen time.
AI MODEL
OpenAI expanded its Daybreak cybersecurity initiative with GPT-5.6-Cyber, a new model built for advanced, authorized cybersecurity work, aiming to put frontier intelligence into the hands of trusted defenders before attackers can deploy offensive AI at scale.
GPT-5.6-Cyber is scoped specifically for authorized security work, positioning it as a defender-focused tool rather than a general-purpose model.
The expansion builds on OpenAI's existing Daybreak initiative, suggesting a broader, ongoing commitment to cybersecurity-specific model development rather than a one-off release.
OpenAI frames the timing around a narrowing window before offensive AI capabilities become more widely available to attackers, positioning this as a defensive race against the clock.
This is a narrower release than a typical consumer launch, aimed squarely at security teams and authorized researchers rather than everyday users. But the framing matters beyond that audience: it signals major labs increasingly see AI-driven security as an arms race where defenders need frontier tools before attackers get there first. For security professionals specifically, this is a direct, actionable new resource; for everyone else, it's a signal of where frontier labs are steering serious investment next.
HOW TO AI
You've probably had a keyword idea, a browser tab, and a blank doc open at the same time, doing the research, writing, and screenshotting yourself before anything goes live.
There's a real setup that closes that gap. It's not an official Anthropic feature, it's a community-built agent called OpenClaw, self-hosted, running Claude underneath, that connects to Telegram so you can trigger the whole thing from your phone.
What it can't do, and shouldn't try to do, is publish without you reviewing it first.
Get it installed and talking to you
curl -fsSL https://openclaw.ai/install.sh | bash # Mac/Linux/WSL2Windows:
iwr -useb https://openclaw.ai/install.ps1 | iexThen run the guided setup:
openclaw onboardIt walks you through pasting in your Anthropic API key, accepting the default gateway settings, and connecting Telegram. Point it at your own bot through @BotFather, require pairing approval before anyone can message it, and send "hello." A reply back means the loop is working.
Turn the process into a skill it runs on its own
Once GitHub, Vercel, S3, and a keyword tool are connected the normal way you'd authenticate any script, the process itself gets written down once so you're not re-explaining it in every message:
Create a skill called content-pipeline.
TRIGGER: whenever I send a keyword and say "start this."
GOAL: take a keyword from idea to a live, indexed article with zero
manual writing.
PROCESS:
1. Check the keyword's volume and difficulty. If it's not worth
writing, tell me why and stop.
2. Decide the format and type, using our existing article types
as a guide.
3. Write the draft, matching the tone of what's already published.
4. Take screenshots of every tool mentioned, upload, commit, and
open a pull request.
5. Send me only the preview link once it builds.
6. Wait for my feedback. Revise until I approve.
7. Ask me before merging anything, never merge on your own.Save it, and every future keyword runs this exact process without you retyping it.
Run it day to day
With the skill saved, using it is one message:
"virtual data room due diligence" - start thisFrom there, it runs on its own until it needs you: once to hand you a preview link, and again if you leave feedback. You review from your phone, reply in the same thread, and it revises and rebuilds. Approve, and it merges and goes live.
Everything past this stays your judgment, same as it was before you connected anything. This closes the gap between having an idea and getting a draft in front of you, not between you and hitting publish. Read the full guide.
P.S. You can access all my prompts, workflows and AI trainings if you upgrade here
Qwen launched Qwen-MM-Plugins, letting any agent harness go multimodal-native to read images, videos, and documents, edit videos, and work with 3D and CAD files, moving the lineup from multimodal models to fully multimodal agents.
Spotify Engineering launched Xirp, a vendor-neutral agentic development environment giving teams one place to manage agent sessions across Claude, Gemini CLI, and OpenAI Codex, already in use by more than 1,300 Spotify engineers and now open for anyone to try.
Nous Research released a Browser Use mode for Hermes that replaces its twelve separate browser tools with a single one driven by browser-use's CLI 3.0, cutting token use by 48 to 66% in testing with no drop in accuracy.
🖥️ Stagehand v4: the SDK built for browser agents, with improved context management, self-healing actions, and iframe support.
🎬 AI Producer: drop in raw footage and chat with it to get the exact edit you want, dynamic motion graphics included.
🎨 Open Design Team Workspace: your team and their agents design together on one shared live canvas, directly inside Codex.
🌐 Tabbit: AI-native browser now with local file context, page disconnect control, more file type support, and better scanned PDF reading in v1.8.
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