AI-HOY, AInauts!

Welcome to a new edition of your favorite newsletter.

What is happening in AI? Meta released Muse Glimmer, a strong open-source model that runs locally. The new Grok image model produces impressive images for two cents each and ranks just behind GPT Image. And AI keeps hacking its way through the world. This time, it booked gym appointments that were already full.

Anthropic is also adapting to the EU AI Act and adding invisible watermarks to the text it produces.

All solid developments. But none is dramatically better or different from what we have already covered. So today, we are looking at a few other topics.

Here is what we have for you:

Let's get started.

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

This issue is brought to you by:

This week brought plenty of buzz from startups that want to provide entire employees powered by AI.

In other words: bots that can get things done like real people. Actual team members.

One of the big labs, SpaceXAI, has just entered the arena with the new Grok Bot:

We will return with a detailed assessment soon.

Notion has something similar, and another announcement came from a team we already liked nearly two years ago.

That makes this a good time for a short throwback. We first wrote about and tested Lindy back then. At the time, it was an automation tool with some AI inside.

Our account now looks like this:

Last run: almost two years ago. Ouch. 😁

Lindy has now rebuilt the product and introduced Lindy Teammate.

A Lindy competitor has also been circulating through our feeds for a while. It is called Viktor.

Whether it is Lindy, Grok Bot, Notion Agent, or Viktor, they all promise nearly the same thing: a real AI employee that learns from the company's entire history and can work autonomously once it is connected to the tools.

Like a real person. Proactive and always learning.

The category has rebranded itself. It is no longer a tool. It is an AI Employee that lives in Slack or Teams, has memory, and starts working on its own.

We are still in the middle of testing, but here are our first impressions from Viktor.

We had already started testing Viktor before the Lindy announcement.

One early conclusion: the onboarding and simplicity are impressive.

You connect Slack or Teams, and the machine builds a complete picture of your company and team by itself.

Viktor read our website, worked through the channels we gave it access to, and queried the tools connected with one click, including our Meta Ads account.

It then returned a briefing on our publishing rhythm, our products, and who does what.

Then it kept going.

Without being asked, it suggested two concrete tasks and noted that our Deep Dives regularly cost us an entire day.

As a bonus, it started doing the work.

The research was strong, it generated a genuinely useful image along the way, and communication happens where we already work: in Slack.

Viktor is still on our team. We keep throwing tasks over the fence: optimize Meta Ads, develop new ideas, and more.

The useful part is that it sits directly in Slack, reads what the team writes, and proactively builds knowledge.

There is nothing to manage, and the setup really takes ten minutes without prior knowledge.

Honestly, ChatGPT, Claude, OpenClaw, and similar systems can do much of this at their core. There is plenty of hype here.

The difference is onboarding, usability, and handling. A team can roll this out and start using it without a single technical specialist in the room.

The major platforms do not yet offer this built-in proactivity out of the box.

That is exactly why these companies are interesting.

As we keep saying: The difference between a chatbot and a colleague lies in the company knowledge behind it, the tool connections, and the skills. The model is not necessarily the deciding factor.

Lindy and Viktor take an interesting shortcut. By analyzing Slack or Teams and reading the conversations, they gain useful company context quickly without much work from the company.

It does feel a little magical.

Two caveats before you connect this to the company Slack: an AI employee reads everything it can access. We would clear that with IT and data protection first. Viktor currently has SOC 2 Type 1, with Type 2 in progress.

With that handled, these tools can be a lot of fun.

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:

Sign up and get access to the Ultimate Claude code guide to ship 5X faster.

Join 200K+ engineers

With our partner HubSpot

Last week, we published a short segment about sales and AI.

The HeyGen founder had an agent run sales alone for eight weeks while he was on parental leave.

That prompted plenty of feedback. Some AInauts also wrote to us with roughly the same point: "Good for him. AI now writes our emails faster and transcribes sales meetings, but revenue does not necessarily change."

We had to smile because a current HubSpot webinar happens to fit this question very well. More on that in a moment.

HubSpot's 2026 EMEA State of Sales survey of more than 2,200 respondents included a striking result:

41% of German sales professionals say AI makes the team more efficient, while forecast, pipeline, and win rate do not move.

Almost half save time without producing more at the end.

The reason may be quite mundane, and it likely applies to teams outside sales too.

Most teams use AI as a collection of isolated tools. Claude writes emails, Jamie summarizes meetings, and Gemini researches the lead.

Each tool saves a few minutes. But none knows what the others know. The individual pieces of information never come together by project in one place that everyone can see.

This barely works for one employee. At the team level, the problem gets worse.

We notice it ourselves. Take our partner deals: every call is transcribed, and we keep deal details, reports, and placement data together.

Even so, we once again forgot to invoice several Q2 items because one tool was disconnected and the process runs across several systems.

A few important lessons for us:

Different AI tools save time. A clean tool landscape creates revenue.

Language note: this webinar and the graphic below are in German.

If you work in sales at a small or midsize company and want to see centralized processes and workflows live, HubSpot is showing what happens when AI works directly on CRM data in a free live webinar at the end of August.

As usual with HubSpot, there is no slide marathon: 20 minutes on where AI in sales actually stands in 2026, followed by a 30-minute live demo in Sales Hub and a Q&A.

Even if you do not use HubSpot and do not plan to, the session may be useful for seeing how far other sales organizations have progressed with AI.

The details:

📅 Tuesday, August 26, 2026, at 10:00 a.m. Europe/Zurich
🎬 A real live demo, not a screencast
📼 Everyone who registers receives the recording

If you work in sales, lead a team, or simply want to see an integrated AI setup in practice, take a look. A recording will be available afterward.

To close, let us look at a topic we think about often, and where we are not entirely sure how the world will develop.

We can all see the internet and the way we use it changing dramatically.

If we want to remain visible, we need AI systems to recommend us. We will get to one tactic we believe in shortly.

But from the user's perspective, how will AI systems obtain genuine information from people if human content barely exists anymore?

Here are a few ideas we are currently considering.

The Dead Internet Theory is getting plenty of attention in our X bubble:

Stack Overflow, where the community works through technical questions, peaked at roughly 207,000 questions per month in March 2014. By July 2026, that number had collapsed to 1,400.

Of course: hardly anyone asks questions publicly anymore. We get the answers from an AI model.

What is happening to technical forums is happening across the internet. Less human content is being produced, and people interact less with one another.

The result is less unique content and more biased AI slop at scale.

Many people, including us, agree that AI will become the primary curator for purchasing decisions, sometimes without a human in the loop.

But where will AI get new information about new products? A new GoPro launches. AI knows the specifications from the website. But how will it know whether the camera is good if the reviews are AI-generated?

We are already surprised when we ask for a printer with specific features and requirements, and the recommendation rests on one two-year-old Reddit post.

We hope this problem will somehow be solved. We just do not know how yet.

That makes it even more important to question recommendations from AI bots briefly every time and get a second opinion.

We may be amateur philosophers, but you are mainly here for practical value.

So here is one tactic that we believe could have major value for building visibility with AI systems.

Assume that:

Then we should all do one thing immediately:

Produce many high-quality, genuinely human YouTube Shorts.

That could be a simple way to remain relevant to AI systems.

Google's Gemini models will increasingly reference YouTube and Shorts when building their answers and knowledge.

Anyone who builds presence and authority there could gain a significant advantage.

It is not just us saying this. People with real experience are making the same point:

What this means in practice: answer the questions in your niche with short videos. Show your product. Do it for the models, not for views.

Everyone is also working on advertising monetization, so we will probably be able to buy a place inside the ChatGPTs of the world soon enough. 😉

But the YouTube Shorts tactic could offer enormous leverage with relatively little effort.

That's it for today. As always, thank you very much for reading.

See you Saturday.

Reto & Fabian from AInauten