Hi AInauts,

Welcome to the latest issue of your favorite newsletter.

We could spend today talking about how much richer Elon Musk just got and how the AI company Cursor was reportedly bought for 60 billion dollars. But that is not really the point today.

Tiny fun fact anyway: technically, your net worth is closer to Bill Gates' than Bill Gates' is to Elon Musk's.

But the bigger story is still the U.S.-government-triggered shutdown of Anthropic's Fable 5 model.

After Monday's issue, a lot of questions and follow-up topics came in.

So today we want to pull the whole thing together under one theme: independence.

Because this is not just an Anthropic story. It is a broader issue that will hit all of us in some way. As always, we are looking at it as practically as possible.

Here is what we have for you:

Let's go.

They optimized for LLMs, not just Google.

FAQs. Comparison pages. Transparent pricing. LinkedIn presence. These aren't vanity plays. They're what gets you cited in ChatGPT, Gemini, and Claude when your buyers are researching, your investors are looking, and your future hires are deciding where to work.

Download the free AEO Playbook for Startups from HubSpot and get the exact checklist. Five minutes to read.

Unlock the playbook

The whole Fable 5 story is genuinely fascinating.

Here is one possible view through the lens of the U.S. government, via David Sacks.

This post showed us just how meta the whole topic has become:

The moral of this fable is pretty clear to us: set yourself up to be as independent as possible in AI.

That is why we pulled together our different approaches again, in a more concrete way.

We see independence in three stages.

You already know stage 1 from us: Files over Tools.

Your knowledge lives in folders, not locked inside one app. That means you can move from Claude to Codex to Antigravity without losing the work.

Short version: work with AI in local folder structures. Let it document progress there. Keep projects in project folders. Out of endless chats, into local Markdown files.

Make sure your knowledge stays with you, not inside someone else's tool.

But what happens when the problem is not one annoying lab, but rising subscription prices, canceled access, or blocked models?

This is where OpenRouter comes in.

In three sentences: one account, one key, access to hundreds of models from more than 75 providers, both closed and open. Claude, GPT, Gemini, MiniMax, all in the same dropdown.

You top up credit instead of committing to another subscription.

The point: you need fewer subscriptions and become less dependent. For almost every model, there is now an almost-as-good alternative.

We already wrote about how OpenRouter's new Fusion model gets close to Fable 5-level performance. Use that for your more complex work.

OpenRouter also has a leaderboard. The current favorite as a daily driver and cheaper alternative to frontier models is MiniMax M3.

Another important piece of independence at this stage is dependence on the providers' own apps.

Claude Desktop with Claude Code/Cowork or OpenAI's Codex are seriously powerful. But without an account there, you also lose access to the app experience.

But we have a very interesting alternative below.

The final stage is the most radical: models that run entirely on your own computer. Offline, private, free.

We have already written about how to check what your computer can actually handle.

Here is the easiest app for getting started with local models: LM Studio.

One app, no Docker fiddling. You download a model and chat locally. With an official plugin, you can also connect your OpenRouter key in the same window.

That gives you local and frontier models side by side.

If your machine can handle it, we recommend Google's Gemma models.

But we should be honest: local models are not at Fable level. The really good ones still need a powerful machine, and they are still quite far away from the top models.

For private, sensitive, simple, offline work, though, they are unbeatable.

Independence does not mean replacing everything with free tools.

It means never again depending on a single provider.

With Files over Tools, OpenRouter, LM Studio, and the right model choices, that is absolutely possible.

You usually have to dial down your expectations compared with the top models, but for many tasks that trade-off is perfectly manageable.

This week we got a question: what do I do with my old Fable 5 chats that I can no longer continue?

Here is a simple workaround:

You find the name at the top of the conversation.

Claude can then pull the information from the old chat and you can keep working with another model.

AI is in 60% of engineering work, but only 20% can be handed off without someone babysitting the output. Join this live webinar on June 24 (free) to see how top teams are using a context engine to level up.

Register Now

Above, we already teased why it matters to become less dependent on apps like Claude Cowork and Co.

The tool we want to show you today deserves its own section.

Chatting through LM Studio is one thing. But many of us now work differently. The AI reads and writes in our folders and completes tasks step by step.

Cowork-style work, basically.

That is exactly why this open-source alternative is interesting.

OpenWork describes itself as an "open-source alternative to Claude Cowork." It is a desktop app for Mac, Windows, and Linux.

It connects to your local folders, reads and writes files, handles multi-step tasks, and supports skills, templates, and permissions. In other words: the Cowork logic, but through an open app.

Of course, you can also connect MCPs and similar tools.

From a usability perspective, it feels pretty close to the other desktop agent apps.

That is why we keep saying in the AI Employee Bootcamp: learn one app and this way of working, and you can transfer the pattern to all of them.

OpenWork runs with your own OpenRouter key. That is stage 2 from above. And it can even connect to local models through Ollama, which is stage 3.

Your folder stays the source of truth. You choose the model from the dropdown.

Plain English: this way of working then belongs to you. No subscription can take it away from you.

You can download the desktop app directly from OpenWork after creating a free account.

After some testing, we genuinely like OpenWork.

It is a very good alternative and definitely makes you more independent.

It is really not rocket science. Give it a try.

To close, here is something that connects directly to last week's loop topic. A small prompt hack for today.

If you have followed us for a while, you may remember that meta-prompting used to be a big topic in our trainings.

Basically, we used prompts that wrote better prompts for our actual tasks.

Today's hack adapts that idea to the agent world.

First, use the /goal skill in Claude or Codex so the AI writes a clean /goal for itself. You then approve it, and off it goes.

That is meta-prompting for everyone. You are not only delegating the work, you are also delegating the briefing. And the AI often knows its own mechanics better than you do.

Here is a prompt template you can copy directly:

Then you only approve or briefly adjust the goals. And every task starts with a clean briefing you did not even have to write yourself.

That is it for today. A pretty independent issue, all things considered.

Thanks for reading us, independent AInauts.

See you next week.

Reto & Fabian from AInauten