In this issue: The battle for your digital middleman or why AI companies are turning to browsers. Three use cases from Germany. Thomas Benkö from Switzerland’s Blick on treating AI like a cheeky colleague. Plus: How to rename screenshots automatically with Google Gemini from the command line.

What we’re talking about: OpenAI will launch a browser, Perplexity just released Comet, and suddenly everyone’s excited about “chatting with tabs.” The browser wars are back, but this time AI companies aren’t fighting over speed—they’re fighting over who gets to be the middleman between you and everything you do online.

Here’s the thing: if you own the browser, your AI gets access to everything that was off-limits before. Emails, shopping carts, paywalled content. No need to negotiate with platforms or circumvent paywalls or IT departments.

If you migrate from Chrome to Comet, all your logins and bookmarks are already there.

Convenient? Sure. But we’re essentially handing over the keys to our digital existence. Our tabs, docs, internal comms, research becomes filtered, analyzed, and monetized. “AI wants the ad dollars too”, Digiday writes. We save a few copy-paste steps. They lock us deeper into their ecosystem. From Reuters:

See for yourself: To get early access to Comet, you have to pay $200 for Perplexity Max. Alternatively, check out the less agentic Dia from The Browser Company (I have 5 invites, hit me up!) or BrowserOS (open source, Y Combinator backed). The line between “helpful AI” and “surveillance as a service”? We’re here to find out where it is.

What else I’ve been reading:

Three use cases: Germany’s AI for Media Network met in Munich to discuss video and image production. After Nikita Roy reminded us in her keynote that we’re “deep into the zero-click era” where most news searches no longer result in clickthroughs, it was hands-on:

1. Animating history: “Geschichtsgulasch” (literally: history goulash) tells historical stories in vertical videos. Klaus Kranewitter and Marcus Müller from Enrico Palazzo Filmproduktion showed how they use “open” image generators like Stable Diffusion and Flux to come up with consistent illustration styles, create images with the help of ComfyUI and LoRA, and use closed models like Sora, Veo3, and Hailuo for animation.
2. Editing news clips: Joerg Pfeiffer from BR’s AI + Automation Lab showed a workflow for creating short videos. Editors source videos from news agencies, write their scripts, and feed both into Gemini. The AI is prompted to identify the best sequences to match the script. The output is an XML file with a rough cut that can be imported into Adobe Premiere. While the system works, Gemini “sees” seconds instead of single frames, preventing it from making perfect cuts.
3. Daily illustrations: Stuttgarter Zeitung launched exclusive subscriber content requiring distinctive visuals six days a week—without dedicated photo editors. Jan Georg Plavec presented their AI-augmented solution: Editors prompt an AI to generate black-and-white illustrations that are placed on colored backgrounds. A custom GPT helps generate image prompts in the desired style. While AI helps, it’s still the human idea that counts.

Let’s hear it from humans: When Thomas Benkö from Switzerland’s Blick is not working on AI generated summaries or tending to his chatbot, BliKI, he’s watching TikTok.

Previously: Bonnier’s Freja Kalderén, Hugging Face’s Florent Daudens, JP/Politikens Hus’ Sara Inkeri Vardar, and Journalist and Media Advisor Jaemark Tordecilla.

Hands on: Rename screenshots with a description automatically with Google Gemini. From the command line. For free. If you’re like me, you take lots of screenshots. I need something I see on my computer for a presentation. I want to remember something I see on social media on my phone. So. Many. Screenshots. And then what?

Google has launched a tool to access Gemini 2.5 Pro from the command line. It’s called Gemini-CLI, and it comes with 1,000 free requests per day. Its main customers are developers, it has some agentic capabilities to access and write code. It lives in your terminal and can easily access files.

Open a terminal and check for npm and node. If you’re not into coding, you might have to install them:

npm -v
node -v

If you don’t have them, first install Homebrew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Next, this should do the trick:

brew install node

On to Gemini-CLI:

npm install -g @google/gemini-cli

Now, got to the directory you want to work in. I suggest starting with a small number of screenshots to dial in the prompt. I made a folder called “screens” on my desktop and put in seven screenshots. In my terminal, I go to cd Desktop, cd screens, and fire up Gemini:

gemini

You have to log in with your Google account once. Next, try your first prompt:

Look at the screenshots in this directory, and rename them with a brief description. Start the file name with the date and time.

Check

/privacy

in Gemini-CLI to disallow data usage, even users in the free tier should be able to opt-out.

How good is that? And it’s just the first idea. I’m sure, you can come up with a plan to vibecode an app that renames screenshots and start making passive income more use cases to work with folders and files on your computer.

A very good quote from Comedian Asher Perlman on Threads:

This is THEFUTURE.