Wir beobachten, was in der Agentic AI-Welt wirklich passiert – LLMs, RAG, Memory, Agents, Tooling – und kommentieren es in unserem Blog. Pointiert, unabhängig, aus europäischer und deutscher Perspektive.
Context windows are the new RAM — and Anthropic proved it by accident. Plus: Google throttles its own agent hype, Shopify quietly builds the smarter stack, and the German podcast scene writes agentic commerce off.
Context windows are the new RAM — and Anthropic proved it by accident. Plus: Google throttles its own agent hype, Shopify quietly builds the smarter stack, and the German podcast scene writes agentic commerce off.
Anthropic has accidentally published the building instructions for its most important agent. 512.000 lines of TypeScript, 1,900 files, one forgotten line in the .npmignore. What the leak reveals about agent architecture is more instructive than any keynote.
Anthropic has accidentally published the building instructions for its most important agent. 512.000 lines of TypeScript, 1,900 files, one forgotten line in the .npmignore. What the leak reveals about agent architecture is more instructive than any keynote.
Anthropic has done more right in three months than OpenAI did in the whole of last year. And the crazy thing is that it wasn't because of the model, but simply because of focus, enough backbone and a smart advertising campaign.
Anthropic has done more right in three months than OpenAI did in the whole of last year. And the crazy thing is that it wasn't because of the model, but simply because of focus, enough backbone and a smart advertising campaign.
Updates of the week: Shopify makes millions of stores shoppable in ChatGPT. OpenAI discontinues Sora and rebuilds Shopping. Anthropic Launch Watch: Computer Use, Claude Code Channels, Safer Auto Mode.
Updates of the week: Shopify makes millions of stores shoppable in ChatGPT. OpenAI discontinues Sora and rebuilds Shopping. Anthropic Launch Watch: Computer Use, Claude Code Channels, Safer Auto Mode.
Companies have no idea how to measure the productivity of their AI agents. So they measure token consumption. Last week, a developer at OpenAI ran 210 billion tokens through their in-house models - enough text to fill Wikipedia 33 times over.
Companies have no idea how to measure the productivity of their AI agents. So they measure token consumption. Last week, a developer at OpenAI ran 210 billion tokens through their in-house models - enough text to fill Wikipedia 33 times over.
OpenAI has abandoned the checkout. Hardly more than a few dozen out of millions of Shopify merchants have ever sold via ChatGPT. Agentic Commerce failed, curtain closed? Not so fast.
OpenAI has abandoned the checkout. Hardly more than a few dozen out of millions of Shopify merchants have ever sold via ChatGPT. Agentic Commerce failed, curtain closed? Not so fast.
Last week, an AI agent at Meta spent almost two hours ensuring that employees had unauthorized access to company and user data.
Last week, an AI agent at Meta spent almost two hours ensuring that employees had unauthorized access to company and user data.
While you are still wondering whether your product data is machine-readable, Google is building the rails on which AI agents will shop in the future. And nobody in Europe is setting the course.
While you are still wondering whether your product data is machine-readable, Google is building the rails on which AI agents will shop in the future. And nobody in Europe is setting the course.
The week of March 14-20, 2026 sent a clear signal: Agentic Commerce stops being a concept and becomes infrastructure. From protocols to payments to the question of whether we can trust agents at all, this week saw building on all fronts simultaneously.
The week of March 14-20, 2026 sent a clear signal: Agentic Commerce stops being a concept and becomes infrastructure. From protocols to payments to the question of whether we can trust agents at all, this week saw building on all fronts simultaneously.
There is that moment in every technology wave when someone starts to draw a map. Not because everything is clear - but because enough players have emerged that you lose track. Cloud computing was ready in 2009, headless commerce sometime in 2020.
There is that moment in every technology wave when someone starts to draw a map. Not because everything is clear - but because enough players have emerged that you lose track. Cloud computing was ready in 2009, headless commerce sometime in 2020.