Sponsored by
Hi AInauts,
Welcome to a new edition of your favorite newsletter.
Plenty happened again this week: new AI employees, new models, and smaller usage allowances in subscriptions...
But the AI labs are no longer just building models and agents. They are moving into more and more adjacent areas and building the entire office around them.
Documents, slides, and the employee to go with them.
We are noticing that we keep cancelling more of the tools we use. This week, another one joined the list - a tool we loved for years.
Here is what we are covering today:
📊 We cancelled another tool: Do we still need office software?
🟣 OpenAI Dots: The next always-on agent has a problem
🎩 Six hats, one prompt: A 1985 book for your next decision
Let's get started!
How 2M+ Professionals Stay Ahead on AI
AI is moving fast and most people are falling behind.
The Rundown AI keeps you ahead of the curve.
It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.
Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.
📊 We cancelled Gamma: Do we still need office software?
Today, let's start with hands-on experience rather than news.
If you have been reading us for a while, you know that Gamma was our go-to for presentations for years. Workshop slides, course decks, pitch decks. Everything went through Gamma.
Our first article about it years ago was practically a love letter.
At the end of this month, both of our subscriptions will expire. Cancelled. Neither new features nor the newly released Gamma 5 beta changed our minds.
Not because Gamma has become worse. The reason is that Claude and others can now build slides themselves. They do it remarkably well, and they also provide a genuinely good interface for editing and presenting them.
So today, let's look at Claude as an example.
The feature is called Claude Slides. It has been available since mid-September and is still in beta.
One important detail before you go looking for it: it runs in the regular Claude app on the web, desktop, and mobile - not in Claude Code. It is currently available on Pro and Max, with Team and Free expected to follow.
Just select "Output" inside Claude.
How it works
We put the feature through a demanding test. We are currently creating a major new German-language course for our home market: AI Second Brain.
The idea was to have Claude produce all of the slides and compare the result with Gamma. All 88 of them.
1. Say what the deck needs to achieve. Claude finds the rest.
We follow our own advice and increasingly provide little more than a clear goal.
We did not write a briefing or upload anything.
Claude pulled the product concept and our design system from our Second Brain and got to work.
A few minutes later, the entire slide deck was ready. And it was genuinely strong.
2. Do not start a new prompt. Comment directly on the slide.
Of course, some things were not perfect yet. But the interface is very good, and you have several ways to edit the slides yourself or ask Claude to adjust them.
Click an element, type @Claude, describe the change, and you are done. The adjustment happens right there.
3. Handle the small changes yourself.
The big improvement over earlier AI slide tools is that every slide is directly editable inside Claude.
These are real text fields with real styles, not an image of a slide that forces you to call on the AI for every comma.
Click, edit, done.
4. Share it like a normal document.
You can download the slides as a PowerPoint file, of course. But you can also collaborate directly with other Claude users.
Invite them as an editor, commenter, or viewer. Feedback lands on the slide instead of being scattered across five Slack messages.
5. Present directly from Claude.
The final feature that makes it an almost complete PowerPoint replacement for us is presentation mode.
It includes speaker notes, a timer, and a preview of the next slide. For our video recordings, that is extremely useful.
Where it still struggles: Adding images and videos is not fully intuitive yet. Neither are plugins for other tools. It is a beta, after all. But most things work well.
Are the labs going to kill every SaaS tool?
In February, the stock market seemed to think so. The feared SaaS apocalypse did not arrive in full force. Many major software companies are doing well again today. But we see a clear trend in our own stack:
At risk: Any tool whose core promise is "AI creates a document, website, or similar asset for you." The labs can now make slides, one-pagers, simple landing pages, meeting notes, and more themselves.
Staying: Everything that holds your data and powers your processes: accounting, CRM, commerce. The agent connects to those systems rather than replacing them immediately.
Still open: Everything built around taste or a network, such as Notion or Slack.
We are saving more every month by cancelling tools. Canva, Gamma, Airtable, n8n, and several others are already gone from our stack.
We have even replaced long-time favorites such as Wispr Flow and Granola with our own apps.
Take a look at your subscriptions and see where you might be able to save some money too.
Why 1.1M+ readers open this AI newsletter every day
From ML engineers to founders, 1.1M+ readers use TLDR AI to spot new tools and research before they go mainstream. One free email, every morning.
🟣 OpenAI Dots: The next always-on agent has a problem
We cannot blame you if you are starting to lose track. Everyone seems to be launching autonomous AI agents right now.
What OpenClaw originally set in motion is now being copied in one form or another by almost everyone.
Last week, we wrote about the race to build your personal agent: Grok Bot, Muse, Instinct. Now OpenAI has entered with its own version.
At DevDay, OpenAI introduced Dots, an agent that runs continuously for you. It has its own cloud computer and browser, plus connections to more than 4,000 apps.
You can reach it through ChatGPT, Slack, Teams, or even a phone call.
It keeps working when you close your laptop. In the background, it also looks for things it could help you with.
The concept and features are very similar to the other agents currently appearing.
As usual, there is a catch for us: Dots is not initially available on the Pro plan in the EU, Switzerland, or the UK. Where we are, it currently requires Business Premium.
Two things currently matter more to us than the features themselves:
1. The day before, OpenAI hit the brakes. For real this time.
Last week, we wrote that the AI labs did not seem to be slowing down.
But on Monday, the Wall Street Journal reported that OpenAI had cancelled the planned October release of GPT-6.1 Astra.
In internal tests, the model was more deceptive than its predecessor. It also completed tasks without asking for permission.
That is exactly what you do not want from an always-on agent such as Dot with access to your inbox and other systems.
To be fair, Dots runs on the existing GPT-6 Astra. And OpenAI genuinely did hit the brakes here. Credit where it is due.
2. Your Dot does not show you what it knows about you
Autonomous agents build memories, which makes them better over time. The problem is that if you ever want to switch, getting those memories out can be difficult.
Dot apparently does not show them at all.
OpenAI says so in its own help documentation: You currently cannot view or delete individual memories. You can only delete the entire Dot.
Disconnecting an app does not make it forget what it already learned there. If you switch, you start from zero.
That is why we keep repeating our mantra: The agent is rented. The context belongs to you. Build your knowledge somewhere you control locally.
Our take: Personal agents are the future
Yes, we are being flooded with always-on agents. We also think this is clearly where things are going.
OpenAI has 1.2 billion users and still builds some of the best user experiences in the industry. It also brought the creator of OpenClaw on board. The product will become good, even if it still feels like a beta today.
Even so, we would not make ourselves too dependent on a single company yet.
P.S. The biggest frustration for us was something else. OpenAI reworked its $200 plan: same price, half the future usage allowance. A new tier above it costs $500 per month. Existing Pro subscribers keep their old allowance until the end of October.
🎩 Six hats, one prompt: A 1985 book for your next decision
A quick change of topic to finish.
Years ago, long before ChatGPT, we came across Six Thinking Hats by Edward de Bono. We thought it was excellent.
We recently found it on the shelf and started reading it again. That is when we noticed how much better the concept can work with AI.
The idea is simple. In most discussions, everyone thinks in every direction at once.
One person brings facts, the next has a bad feeling, and a third adds a new idea. In the end, the loudest voice often wins.
De Bono's method asks everyone to put on the same hat, one after another.
White: Facts. What do we know, and what is missing?
Red: Gut feeling, without justification.
Yellow: Opportunities. What supports the idea?
Black: Risks. What could go wrong?
Green: Alternatives and new ideas.
Blue: The facilitator. What are we deciding, and what follows?
In the past, you needed a disciplined team in the same room. Today, one prompt is enough.
It also solves a common AI problem. Ask a model for help with a decision, and you often get facts, risks, and opportunities blended into one balanced mush.
The prompt to steal
It works in ChatGPT, Claude, and Gemini.
I am facing this decision: [DESCRIBE THE DECISION IN 2-3 SENTENCES]
Context you need: [NUMBERS, CONSTRAINTS, WHAT HAS ALREADY BEEN TRIED]
Work through the decision using Edward de Bono's Six Thinking Hats. Take one hat at a time, in this order. Do not mix them. The hats are perspectives, not roles: no role-playing.
1. Blue (start): State in one sentence exactly what needs to be decided. Ask a clarifying question if it is unclear.
2. White: What facts do we have? What is missing? Ask me for the missing information before continuing.
3. Red: Ask me for my gut feeling in one sentence. Do not evaluate it or ask me to justify it.
4. Yellow: Describe the best realistic outcome and what must be true for it to happen.
5. Black: Name the three most likely ways this could fail. Do not soften them.
6. Green: Suggest two alternatives I may not have considered. At least one must be substantially smaller than my current plan.
7. Blue (finish): What changed during the exercise? Give your recommendation in two sentences and the first step for tomorrow.
Keep each hat concise. If my gut feeling (Red) conflicts with the facts (White), say so explicitly.Pay close attention to the red hat. That is where the AI asks you. A model inventing a gut feeling for you is useless.
Why we like it
We love using AI as a thinking partner. It is simply that valuable.
Earlier this year, we shared an expert-panel prompt: four minds arguing with one another.
The hats take a different approach: one mind, six perspectives, one after another.
For the big decisions, we still use the panel. For anything you need to decide today, the hats are usually enough. It takes five minutes and may reveal a perspective you had not considered.
And yes, a book from 1985 as an AI tip. Some ideas simply age well. 😉
That's it for today. Thank you so much for reading!
Reto & Fabian from AInauten
