👨‍🚀 Reflexive AI: Why companies need to rethink their approach

Hello AInauts,

Welcome to the latest issue of your favorite newsletter!

This is what we have with us today:

• 🦙 Meta's new Llama models are... bad!

• 🔥 Reflexive AI: Why German companies need to rethink their approach

• ✍️ Our current favorite AI notetaker for meetings from 🇩🇪

Here we go!

🦙 The new Llama models from Meta are... bad!

We are delighted to be able to start with news about new open source language models.

And it's not just that we can finally use the 🦙 icon again. When it comes to open source language models recently, the news has mostly come from China around DeepSeek.

But on the weekend, the time had finally come and Meta presented the eagerly awaited new Llama 4 models.

Scout, Maverick and Behemoth. All three are great multimodal models with insanely large context windows, and they perform slightly better than other models in their weight class.

Scout's 10 million token context window, i.e. its memory, is particularly impressive 🤯 - and Maverick also goes up to a million.

Nerd corner: If you're wondering what the 16 or 128 experts in the screenshot above are all about: Llama 4 is also a MoE model. MoE stands for "Mixture of Experts" - a smart AI design in which a language model consists of several specialist networks (the "experts") instead of a single large network. A MoE model has a router that decides which expert is best suited for each task. The whole thing has efficiency and cost benefits.

Of course, there has been a lot of controversy about the new models online over the last few days. But we don't want to go into it in detail - if you're interested, you can read all about it here.

The fact is: Meta has introduced some great new models - and you can already download Maverick and Scout and use them for free.

You want to try out Llama (for free)?

Sure, of course you can download it. But it's a bit technical and requires fast hardware.

As always, our friends at OpenRouter.ai are fast and already have Scout and Maverick live as a free version.

You can try them out here and compare them with others directly in a chat.

For all automation engineers: Of course, like everything else with OpenRouter, this is also available via API and can be used immediately!

🔥 Reflexive AI: Why companies need to rethink their approach

An internal memo from Shopify CEO Tobi Lütke has just gone viral in the tech world.

It's about the use of AI in his company. And when one of the most successful German entrepreneurs of our time writes something about it to his people, it's worth listening to!

First: If you don't know Tobi Lütke, it's worth reading all the way to the bottom and watching the podcast. It's pretty awesome. Here is the link to the memo:

Tobi Lütke communicated internally (and later publicly): "Using AI effectively is now a fundamental expectation of everyone at Shopify!"

Meaning: The use of AI is no longer an option, but a basic expectation for ALL employees!

In other words: AI is not a nice-to-have goodie or a cool recommendation - it is a clear statement that AI is now part of the standard toolkit. Those who don't follow suit will fall by the wayside.

What does "reflexive AI usage" mean? 🧠

The title of the memo, "Reflexive Al usage is now a baseline expectation at Shopify", is somewhat cryptic at first glance. What it means is that you no longer consciously use AI as a separate tool, but automatically integrate it into every task.

Like a reflex! You write something, plan something, brainstorm something? AI is simply there without you giving it much thought.

Some concrete examples:

• Marketing agencies use deep research to build everything around new customers - and use the result to knit campaigns with AI.

• Founders speak their thoughts directly into Reflect.app and use the transcript to challenge it along with ChatGPT.

• Support teams build CustomGPTs for simple requests - without asking developers for dev support.

• Project managers give all AI-generated meeting notes to Claude and get input for a presentation. They then create it in minutes with Gamma.

The German contrast 🇩🇪

Shopify integrates AI usage into performance reviews. Teams are expected to share prompts and workflows with each other - and that everyone learns AI themselves.

While Shopify is making AI the standard, many companies are still stuck in the "we-should-maybe-test-the-copilot-first" phase. And even that often happens with no end of reservations and restrictions.

To be fair, we have also often heard: "My company needs to pay for my AI training..."

Our take: This reluctance could put German companies at a real disadvantage in the medium term. While teams at all levels in North America and Asia are multiplying their productivity with AI, we are still discussing data protection concerns when using ChatGPT.

Of course, compliance is important! But instead of blocking ourselves completely, we need smart solutions:

• Company-wide AI strategy with clear guidelines

• Internal training on the effective and safe use of AI

• In-house, data protection-compliant AI solutions (e.g. with n8n)

• Performance reviews that take AI expertise into account

Why do we think it's relevant?

Tobi Lütke writes something in his memo that should wake us all up:

"Honestly, I don't think it's feasible to avoid learning the ability to use AI. You're welcome to try, but I'll be honest: I can see that it won't work today, and it definitely won't work tomorrow. Standing still is almost certain, and standing still is failure in slow motion. If you don't climb to the top, you slide down."

This applies to individuals as well as entire companies. Standing still is a step backwards, especially at the current pace of AI development.

What can YOU do?

I mean, you read this newsletter. So you're at the forefront anyway. But maybe a few tips for your colleagues 😉:

1. Become an AI pioneer yourself: use AI reflectively in your day-to-day work, even if your company is not (yet) actively promoting it. Invest the 20 euros in ChatGPT Plus.

2. Share your successes: Show your team and bosses concrete examples of how AI helps you deliver better results.

3. Demand better tools: Make it clear that a slimmed-down version of Copilot is not enough to stay truly competitive!

4. Educate yourself: Get to know the latest AI tools and techniques - and try them out! You're already reading the newsletter. That's great.

Conclusion: A new era in the world of work

The Shopify memo marks the beginning of a new era in the world of work. Companies that only see AI as a nice add-on will be left behind by those that integrate it as a fundamental tool in their DNA.

The question for German companies is no longer IF they should use AI - but HOW FAST they can establish a true AI-first culture.

P.S. Tobi Lütke is a brilliant person, (forward) thinker and entrepreneur. If you want to dive deeper into his thinking, listen to this podcast.

✍️ Our current favorite AI notetaker for meetings from 🇩🇪

Today's issue has become quite long again. So let's finish with a brief tool recommendation:

Recording and summarizing meetings with AI is already cold coffee for many. There are countless on the market. TLDV, Otter, Fireflies... Slack, Zoom and co. have one built in and so on.

BUT we want to briefly introduce you to our current favorite tool because it is a little different from others.

The tool is called Jamie, comes from Germany and is GDPR-compliant.

The big difference, why we love it: You can use it to record and summarize meetings WITHOUT having to bring an annoying bot into your meeting.

It's as simple as having Jamie record and transcribe the audio directly on your computer.

Simply start the meeting and off you go. And yes, you could also use it to record meetings "secretly". But you don't do that.

Of course, Jamie has all the features of the others.

• Summaries

• Chat with your meeting notes

• etc.

A game changer for us: we also use it a lot for offline meetings on the computer.

This brings us to the not unselfish part: unfortunately, Jamie doesn't have a mobile app yet. But it would be brilliant for offline meetings! It's also by far the most requested feature on the roadmap.

Maybe you can try Jamie here for free. And if you like it, upvote it here. The more upvotes, the better. We want a mobile app. 😉

We made it! But no need to be sad. The AInauts will be back soon, with new stuff for you.

Reto & Fabian from the AInauts

P.S.: Follow us on social media - that motivates us to keep going 😁!
X, LinkedIn, Facebook, Insta, YouTube, TikTok

Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.