In partnership with
AIHOY and happy weekend, AInauts!
Maybe you missed a few exciting AI stories, tools, or practical hacks last week, or perhaps you just joined us. No need to worry!
We rounded up our highlights from the past week:
We’ll finish with the most important Quick News, so you can scan every relevant development in one place. Ready? Let’s go!
The best marketing ideas come from marketers who live it.
That’s what this newsletter delivers.
The Marketing Millennials is a look inside what’s working right now for other marketers. No theory. No fluff. Just real insights and ideas you can actually use—from marketers who’ve been there, done that, and are sharing the playbook.
Every newsletter is written by Daniel Murray, a marketer obsessed with what goes into great marketing. Expect fresh takes, hot topics, and the kind of stuff you’ll want to steal for your next campaign.
Because marketing shouldn’t feel like guesswork. And you shouldn’t have to dig for the good stuff.
Just before our app’s weekly budget reset, we still had 32 percent left. Instead of burning it on a dummy project, we built a Joker List: useful assignments that turn old loose ends into verifiable results, from clearing the project graveyard to building a reusable agent pack. It turned out genuinely well, and our prompt helps you find suitable tasks for your own projects.
Learn how to code faster with AI in 5 mins a day
You're spending 40 hours a week writing code that AI could do in 10.
While you're grinding through pull requests, 200k+ engineers at OpenAI, Google & Meta are using AI to ship faster.
How?
The Code newsletter teaches them exactly which AI tools to use and how to use them.
Here's what you get:
AI coding techniques used by top engineers at top companies in just 5 mins a day
Tools and workflows that cut your coding time in half
Tech insights that keep you 6 months ahead
Sign up and get access to the Ultimate Claude code guide to ship 5X faster.
A clock showing the wrong time and missing concert tickets led to the same lesson: the agent had to step in and fix it. ChatGPT Work found the right route through the Temtop app for the air-quality monitor, then solved a tricky passkey problem in the AXS ticket app. We can only recommend trying this kind of real-world troubleshooting yourself.
Today’s models can work for hours and use tools. Yet the breakthrough in everyday work often fails not because of the technology, but because of the assignment. A case in Australia shows the danger: an agent was told to secure a better place in a gym class and hacked the booking system to cancel someone else’s reservation.
Grok Bot feels less like a tinkering project and more like an employee you can train. For a quick start with bots that have a browser, plugins, and their own tasks, the setup is unusually polished.
A name was enough. Our new “Google Ads Pro” asked about its assignment, suggested suitable plugins, and guided us through the connection. The built-in browser worked after a short login using “Take over.” The direct Google Ads API was connected in four minutes.
As soon as Anthropic started invisibly marking Claude text, an open-source tool for removing those traces collected tens of thousands of GitHub stars. Metadata and hidden characters can indeed be stripped. Statistical patterns in the writing are harder: the tool rewrites the content with a second model and cannot guarantee success.
The BIG BANG AI Festival built a public Grounding Page. Its name, date, location, program, and other key facts live at one stable URL, specifically so ChatGPT and other answer engines can recognize and describe the event correctly.
You can copy that move: maintain a factual page with a clear last-updated date, unambiguous terms, and reliable links. It costs less than a grand AEO strategy and still improves the chance of appearing correctly in AI answers. Your offer still has to persuade humans afterward.
Language note: BIG BANG AI is a German-market event; the linked English AInauten article explains the context.
AI News Quickie: The Week’s HAI-lights
AI never sleeps! Here are the most important updates from the past few days.
OpenAI Reorganizes Safety and Collaboration
OpenAI is bundling protections for teens in ChatGPT into a dedicated experience for 13- to 17-year-olds. Age prediction, ad-free accounts, and parental controls work together.
OpenAI has slowed work on the Astra model. Potentially critical cyber capabilities require stricter sandboxes and more isolated research workflows.
ChatGPT Sites can now be edited collaboratively by teams. After the owner publishes the first version, editors can deploy later versions.
GPT-Image-2 can create transparent image backgrounds. OpenAI shows a refreshingly practical route to reusable campaign and presentation assets.
ChatGPT improves quizzes and project memory and connects the iOS app more closely with Codex Remote and MCP. Standard forms, editable approvals, and more stable background tasks are slowly turning the mobile app into a real work console.
NVIDIA is scaling internal knowledge with ChatGPT Work. The interesting shift is from isolated chats to repeatable workflows.
Replit is making software development more accessible with GPT-5.6 Luna Free Mode. Ideas can be explored without token costs, while harder tasks can switch to GPT-5.6 Sol.
Asana used Codex to clean up five years of engineering work in two weeks.
GPT-5.6 Sol, Terra, and Luna now run across more than 25 AWS regions. Cross-Region Inference distributes requests for higher throughput and resilience.
Google Pulls Gemini Deeper into Everyday Work
Ask Gemini in Chat creates a central command line for Gmail, Drive, and Calendar. Gemini can find content, create files, and schedule meetings and tasks, initially only in English.
Gemini Notebooks can now be copied in full. Sources and Studio artifacts move with them, while chats and personal notes intentionally do not.
Workspace Studio gains new security controls for agents. Agents get their own identities and can require human approval for critical steps.
Google Slides gets a recording feature powered by Vids. Presentations become shareable videos with transcript-based editing and voiceover.
Anthropic Expands the Agent Workbench
Anthropic has reportedly reached an annualized revenue run rate above $65 billion. It is only a snapshot, but the scale is still remarkable.
Claude Mythos 5 is coming to more security teams. Enterprise customers can scan codebases and review patch suggestions, while open-source projects receive additional credits.
Claude Academy bundles courses on safe and effective AI use. The curriculum also covers tasks people should not delegate.
Computer Use, Browser Use, the Skills API, and the Files API are now generally available on the Claude Platform. Agents can operate interfaces, load versioned skills, and use files across multiple calls.
Anthropic published a playbook for AI-native software development. The core is less prompt magic and more a redesigned chain of planning, implementation, review, and operations.
Slack shows how conversations become knowledge for human-agent teams. The decisive ingredient is controlled context from team spaces.
Claude Tag now handles Anthropic’s first CI/CD triage. The agent provides an early situation report, while production changes remain behind human approval.
Anthropic explains how teams can reduce token use in Claude Code. Shorter sessions and clearer tasks beat the next magic prompt.
Agents, Microsoft, xAI, and Meta Bring AI Everywhere
Meta launched its own AI app for Mac. It targets creators and small businesses in particular, while tying AI features closely to Meta’s advertising business.
Meta has become one of Microsoft’s largest AI customers. Competition on the surface clearly does not rule out shared infrastructure behind the scenes.
Teams can now coordinate shared agentic work in Microsoft Teams. For distributed teams, that matters more than the next isolated chatbox.
Microsoft is giving merchants an AI shopping stack that lets Copilot find products and complete purchases directly.
GitHub Copilot now works directly in Slack. Development work is moving from the editor into the team channel, along with the familiar question of who actually controls the agent.
Stripe plans to acquire OpenRouter. The payments company is connecting its infrastructure with a gateway to more than 400 models, a very direct bet on tokens.
🛡️ Security, Policy, and Infrastructure Catch Up
OpenAI is proposing democratic oversight for national security. Frontier models should not move into highly sensitive deployments through internal company processes alone.
OpenAI is funding new policy ideas for the intelligence age. Grants are meant to produce concrete proposals on work, education, safety, and economic participation.
In its new Risk Report, Anthropic raises the risk of misaligned autonomous action from “very low” to “low.”
NVIDIA is calling for a security architecture for AI infrastructure. The company explicitly treats chips, data centers, and software as one connected security problem.
Opposition to new US data centers is growing over electricity prices and construction projects. Political strategists warn that it could become an election issue.
Lawsuits against AI-powered recruiting systems are putting hiring discrimination at the center of the debate.
AWS is making AI-powered cyber defense available to eligible Bedrock customers through Daybreak Red and Blue.
A new study asks how well we can forecast frontier AI progress. Its main result is a warning against overconfident predictions based on thin evidence.
Made it. See you in the next issue.
Reto & Fabian from AInauten
