Hello AI Enthusiast,
This week, AI is moving into the tools we use every day. Anthropic gave Claude its own identity inside Slack, OpenAI taught Codex to learn by watching your screen, and Midjourney surprised everyone by building a body scanner that looks like a spa. Let's dive in.
The Big Picture 🔊
Claude Comes to Slack as a Team Member!
Anthropic has launched Claude for Slack with a new agent identity system that gives Claude its own workspace-level permissions, separate from individual user accounts. Each channel gets a distinct Claude identity with its own memory and access rights, so what Claude can do in one channel stays separate from another. Admins control what tools and integrations Claude can access across the organization.
|
||
|
This is different from just adding Claude to your Slack workspace. When you tag Claude in a channel, it's not your personal Claude with your Gmail, it's a shared assistant that only has access to what your admin set up for the team. Think of it like a new colleague who has access to shared team tools but not your personal inbox. The practical upside is you can ask Claude to pull from Granola notes, update a Notion page, or help with a live discussion, all without switching tabs. The catch is it needs to be set up properly. Give it the wrong permissions and either it can't do much, or it can do too much. |
OpenAI Codex Can Now Learn by Watching Your Screen
OpenAI has added a record-and-replay feature to Codex that lets users record themselves completing a task on screen, then have Codex turn that into a reusable automated skill. The idea is to make automation as simple as showing the AI what you do, rather than writing instructions from scratch. Currently available on macOS only, with the European Economic Area, UK, and Switzerland excluded at launch.
|
||
|
The concept is genuinely interesting: show it once, it does it forever. In practice, it still runs into the same wall most browser automation hits, which is anything behind a login or requiring manual authorization tends to break the flow. Token consumption is also high for longer tasks. That said, this is moving fast. A few months ago the same approach in Claude for Chrome already worked well enough for simpler workflows. The geographic and OS restrictions at launch are frustrating, but they won't last. Worth watching. |
Midjourney Is Building a Body Scanner That Looks Like a Spa
Midjourney, best known for AI image generation, has announced it is developing an ultrasound-based body scanning device. The scanner submerges users in water and uses electrical currents and AI to reconstruct a 3D image of the body, aiming to make detailed health scans as accessible as a routine wellness visit. The company, which has no outside investors and runs on $500 million in annual revenue with around 100 employees, is framing this as a research project backed by its community.
|
||
|
On the surface, this looks like a strange pivot. Midjourney went quiet after dominating the early AI image generation wave, and now they're building medical hardware. But look at the numbers: $500 million in revenue, 100 people, no investors, and a team that has already solved a hard technical problem in generative AI. They have cash, talent, and nothing left to prove in their original market. The technology is not as far from what they know as it seems. Turning one type of signal into a high-quality image using diffusion models is basically what they already do, just with electromagnetic data instead of text. Whether the spa aesthetic lands or not, the underlying bet on preventive health and body monitoring is tapping into a market that is genuinely growing fast. |
Bits and Bobs 🗞️
Google's Interactions API is now generally available, offering stable, stateful AI agent workflows and becoming the default interface for Gemini models and AI Studio.
ChatGPT Enterprise now offers detailed usage analytics and flexible spend controls to help admins manage AI deployment and costs more effectively.
Mistral OCR 4 is a powerful AI model that extracts structured text with bounding boxes, classifications, and confidence scores from documents in 170 languages, supporting fully self-hosted deployment and advanced document workflows.
OpenAI's new GPT-5.5-Cyber model and Codex Security tools help defenders automatically find, validate, and patch software vulnerabilities to improve cybersecurity defenses.
Claude Code now creates live, shareable web pages called artifacts that automatically update with your session's AI-driven work progress for easier team collaboration and real-time status sharing.
From Our Founder’s Channels 🤳
In his latest TikTok, Gianluca Mauro reveals who may have snitched on Anthropic. Watch it here.
@gianluca.mauro Who snitched on Anthropic? #Ai #learnontiktok #business
And in his latest LinkedIn post, our founder Gianluca Mauro shares the ALEC Holdings approach as a great way to set up your AI transformation. Read it here.
It's easy to read about a company getting AI right. Doing it inside your own team comes down to the right approach and the right skills. Our Corporate AI Training programs help your people build both.
LOLgorithms 😂
That's a wrap on our newsletter! Before you go, here’s a quick recap of our offerings:
AI Academy Membership: Get 12 months of access to all our cohort-based programs and on-demand courses.
AI Agent Bootcamp: Accelerate processes and solve business problems by building AI Agents, without coding.
Practical Introduction to ChatGPT: A free course on using ChatGPT confidently, understanding its workings, and exploring its potential.
Customized Corporate Training: Equip your team with the skills they need to unlock the potential of AI in your business.
Catch you next week! 👋
