Good morning! Here's what's happening in AI today:
Meta launches Muse Code, a terminal coding agent for long-horizon work
Grok brings Connectors to Voice Mode, ask about emails just by talking
Slack launches Slackbot Surfaces, Slides, Code Execution, and Image Generation
How to turn an outdated local website into a paying client, step by step
4 new AI tools worth trying today
AI CODING
Meta introduced Muse Code in beta, a terminal coding agent built for long-horizon software engineering and powered by its new Muse Spark 1.2 model, which plans, implements, and validates complex changes across large repositories.
Muse Code works across multiple files at once inside large repositories, rather than handling one file or one function in isolation at a time.
It runs persistent sub-agents that solve difficult problems with less back-and-forth, so a task that used to need constant check-ins can run further on its own.
The whole thing sits in a terminal, not a separate IDE plugin, aimed squarely at engineers who already live in the command line.
Most coding agents still stumble once a change spans more than a couple of files, this one is built specifically for that harder, multi-file case. If persistent sub-agents genuinely reduce how often a human needs to step in, that changes what "long-running" actually means for an agent, hours instead of minutes. It's beta software, so the real test is a real repository, not the demo.
AI TOOL
xAI rolled out Connectors inside Grok's Voice Mode, letting people ask about their emails, check daily meetings, or use any connected service just by talking instead of typing.
Voice Mode can now reach into whatever's already connected, email, calendar, and other linked accounts, without switching over to the text interface first.
This turns Voice Mode from a conversational novelty into something that can actually answer questions grounded in a person's real data.
It works with any existing Connector already set up, so nothing new needs to be configured beyond what's already linked.
A voice assistant that can only chat is a party trick, one that can check your actual calendar or inbox while you're doing something else with your hands is a real tool. This closes the gap between Grok's voice interface and its text-based capabilities, they now do the same things. The open question is whether people trust a voice-only answer about something as specific as a meeting time, without a screen to double-check it against.
AI TOOL
Slack launched Slackbot Surfaces, Slides, Code Execution, and Image Generation together, aiming to make AI fluency the default for every employee rather than something only the technical few reach for.
Slides lets someone build a presentation directly inside Slack, without opening a separate app or exporting anything.
Code Execution means Slackbot can actually run code and return real results, not just talk about what code might do.
Image Generation puts visual creation in the same place people already spend their day, instead of a separate tool with its own login.
Most workplace AI tools still assume some baseline technical comfort, this bundle is explicitly aimed at closing that gap for everyone else on the team. Putting slides, code execution, and image generation inside the tool a company already lives in removes the friction of picking which app to open for a given task. Whether non-technical employees actually adopt it, or it becomes one more feature nobody discovers, comes down to how well Slack surfaces it.
HOW TO AI
Somewhere within twenty minutes of you, a business with great reviews is losing customers over a website nobody's touched since 2013. Not because the business is bad, because the website is the first thing anyone sees, and right now it's telling people to go somewhere else.
Cold pitches make people imagine something, and imagining is where "no" lives. Build the site first. Show up with it already done, and the whole thing turns from a pitch into a simple yes or no.
Four tools run the whole system, start to close, all on free plans.
Find businesses a bad website is actually costing
Open Google Maps, search narrow ("dentists in West Boulder," not "dentists in Texas"), skip the first three results since they usually already have an agency. Scroll to businesses ranked 4 through 20. The target: live on Google 5+ years, under 40 reviews, a site frozen in 2013 or no site at all, reviews good despite it.
Collect 35. Name, phone, current site link, one specific detail about each, that detail is what gets your message read instead of reported.
Let Claude write the diagnosis, and let Lovable build the site
Paste the list into Claude with a prompt asking for a diagnosis, a site brief, and a cold outreach message per business, three separate outputs, not one blob. Then feed the brief into Lovable:
Build a landing page for [business name], a [type of business] in [city].
Audience: [who their customers are]. Brand feel: [3 words from the brief].
Hero focus: [the main thing from the diagnosis].Claude Code fetches the source material on its own, then Graphify does the real work: reading every page, extracting the actual ideas inside them, not just filing pages as files. One run on the Claude Code docs turned 145 source documents into 591 separate ideas, wired together by 685 connections, sorted into 67 topic groups.
Turn it into a video, then send it without ever mentioning AI
A screenshot reads as marketing. A video of their own site moving reads as made for them, and it roughly triples reply rates. Upload 3-5 screenshots to Higgsfield with a 10-second cinematic walkthrough prompt, vertical, 9:16, not horizontal.
Send it with a message that never says Claude, Lovable, Higgsfield, or "AI," business owners have been pitched enough $500 chatbots this year to flinch at the word. Then close on a call, not over email, the moment someone's looking at their own site on a screen-share, they're not deciding if they want a website, they're deciding if they want this one. Read the full guide.
P.S. You can access all my prompts, workflows and AI trainings if you upgrade here
Visual Studio Code shipped built-in dictation, letting developers code by voice at what the team calls the speed of thought, no third-party extension required.
v0 launched a new API that gives programmatic access to app building, so developers can ship their own app builder, hand an agent a way to build and deploy, or generate apps straight from a script or CI job.
Prime Intellect introduced Prime Agent, a self-improving harness for coding and long-running autonomous tasks, built to be token-efficient through programmatic tool calling, context as a variable, and multi-agent messaging.
🎨 Open Design: Codex Design is now live in the official Codex Plugin Directory, bringing a real design canvas directly into Codex.
🌐Hark: Handoff can order food, book flights, shop, and deeply research, all through natural language, more affordably than frontier browsing models.
📸Reve: Reve on iOS is out now, transform everyday photos into something fun, polished, or business-ready right from your phone.
🧠MemoryPlugin: Now syncs local Claude Code, Codex, and Cursor sessions into the same searchable memory as your ChatGPT, Claude, and Gemini chats, one shared history across 21+ AI tools.
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