👨‍🚀 How to make ChatGPT think like a genius

Moin AInauten,

Welcome to the latest edition of your favorite newsletter!

GPT-5 has us hooked. We spent two days researching, testing, cursing, tweaking, and optimizing. And in the end, we distilled the results into the prompt tips and updates below.

Here's what we have in store for you today:

• 🔥 GPT-5 News: Important updates from the last 48 hours

• 💭 5 Prompting Hacks for ChatGPT You Should Know About

• 🙀 Could AI wipe out humanity in 10 years?

Let's go!!

🔥 GPT-5 News: Important updates from the last 48 hours

Wow... if you believe the comments on X, Reddit, and Co., then GPT-5 was a complete bust.

But keep in mind: status quo bias and loss aversion make new things seem suspicious at first. Only routine makes them useful. ... we still remember well the negative reactions when Facebook introduced the news feed - now this is the core feature of every social network.

Is the volume of criticism therefore perhaps less a judgment on GPT-5 itself, but rather an echo of a general aversion to change? …

Sam commented on X: "We expected some bumpiness as we roll out so many things at once. But it was a little more bumpy than we hoped for!"

This makes it all the more important for OpenAI to take user feedback and criticism seriously and address it with appropriate measures. Let's take a look.

The main points of criticism – and OpenAI's feedback

• Many users have vented their frustration about the "stupid model" right after the launch. But the problem wasn't with the model at all: there were a few issues during the rollout.

• For example, the router that was supposed to forward the request to the appropriate model did not function properly. Complex requests were then forwarded to a simple minded GPT-5 version instead of one that could properly consider them and provide appropriately good answers.

• This routing problem has now been fixed, so the responses are better!

• Another major criticism was that all previous models were immediately replaced and were no longer accessible. Many users had developed an emotional attachment to GPT-4o, which was abruptly severed.

• OpenAI has therefore reintroduced the 4o model for paying users, at least for a certain period of time. However, you must explicitly activate it in the settings!

• Bonus: Here you can take a blind test to see whether you prefer the answers from GPT-5 or 4o.

• Users criticize that GPT-5 is less creative and experimental than its predecessors, which is particularly evident in role-playing games and creative writing.

• Specialized workflows and custom GPTs also no longer function as expected, as GPT-5 responds differently to familiar prompts.

• Sam has therefore announced that the personality of GPT-5 should be made "warmer" soon (although the writing skills are impressive).

• The initial usage limits for Plus users were set at 200 messages per week for Thinking Mode. Free users reach their limits after just a few messages.

• According to Sam, the usage limit for paying Plus users will now be increased to a whopping 3000/week – nice! And maybe we’ll even get a few Pro requests on top of that.

• There were also expectations that a new image model and the Sora 2 video model would be incorporated into GPT-5 for stronger multimodal capabilities. Unfortunately, this has not yet happened…

On the other hand, we should also appreciate the scope of this mammoth project!

• OpenAI has not only collaborated with Microsoft, but also with Google and other major players...

• And, of course, we must also commend the OpenAI employees who made this possible.

• Sam knows this too, which is why he is handing out multimillion-dollar extra bonuses to many key players for not defecting to Meta (or others).

• Plus: soon you will also be able to cash in your shares sell them!

• However, the most important thing from our perspective is that OpenAI is not only raising the ceiling but also the floor.

• In other words: 700 million users worldwide will have access to the best reasoning model for free! And that will change every conversation, every document, every decision.

Well, that's it for the most important updates since the launch... Let's now turn to the exciting question of how you can get the most out of GPT-5.

💭 5 Prompting Hacks for ChatGPT You Should Know

OpenAI has published a comprehensive guide and a prompt optimization tool for GPT-5. This allows us all to learn how to use it most effectively.

Hack #1: Force (or stifle) Thinking mode

As mentioned, GPT-5 has an internal router that automatically chooses between a fast standard model and a slower but more thorough Thinking model.

The problem? This router sometimes makes overly conservative decisions. However, you can use specific trigger words to make the system think for itself.

Use phrases such as:

- Think about it step by step and in depth.
- Take time to analyze all the nuances before responding.
- You MUST plan carefully; the outcome depends on it!

Pro tip: You can also skip Thinking mode at any time by clicking on “Get a quick answer”.

Bonus tip: Always open new tabs as ChatGPT thinking windows.

We have already experimented with Custom Instructions to activate the Thinking Model more often. However, there is an even simpler, more efficient solution...

To ensure that you use ChatGPT more often, you can also automatically add ChatGPT when opening new tabs. To do this, use a browser extension such as "Custom New Tab" (Chrome, Edge, etc.) and enter the following URL there.

This URL contains the Thinking Mode parameter, which opens a new chat in Thinking Mode directly. Note: Currently only works for paying users.

https://chatgpt.com/?model=gpt-5-thinking&q=

Hack #2: Advice from the experts

This is one of our favorite hacks! Instead of getting a single answer, you let GPT-5 simulate a debate between several experts. This integrates multiple perspectives into a single prompt. Here is a simple version of the prompt:

## Expert panel

### Analysis
Put together a diverse team of experts with critical thinking skills. Review assumptions and special cases in dialogue, ensuring consistency and diversity. Decide on further dialogue or fine-tuning.

Use the following XML tags in the single code block:
<expert name="STRING" field="STRING">Insight</expert>
<speaks name="STRING">Answer</speaks>
<draft version="STRING" by="STRING">Iteration</draft>
<feedback by="STRING" on="STRING">Feedback</feedback>
<revision version="STRING" by="STRING">Revision</revision>

Briefly summarize each step and the overall process outside the block.

### Output
Response based on the analysis as continuous text. Use lists or tables sparingly. No dashes. Paragraphs and structural elements for better clarity.

### Request
{YOUR REQUEST}

Hack #3: Stacking ChatGPT personalities

ChatGPT has also been given new personalities ("Cynic," "Robot," "Listener," "Nerd"). This feature makes interaction more human and definitely more entertaining. And you can also dynamically consult these personas without having to start new sessions.

Okay, now give me the answer from the perspective of your other personalities (Cynic, Listener, Nerd, Robot).

Hack #4: Tree-of-Thought for strategic planning

Instead of linear prompting, you can also instruct the model to explore multiple reasoning paths and then refine the optimal path further.

Explore this using the Tree of Thought: generate 3-5 alternative paths, evaluate the pros and cons, then select the best one and refine it further.

Hack #5: Use the GPT-5 Prompt Optimizer

The new GPT-5 Prompt Optimizer is a super helpful tool, you can find it here!

1. Just paste in your prompt.

2. Add any instructions in the chat box and click on "Optimize."

3. Then click on the icons and see exactly what has been optimized in the background.

These prompts aren't just neat tricks. They're your strategic asset!

Test them and integrate them into your regular workflows. The future belongs to those who not only ask better questions, but also design and control smarter systems.

We love diving into these rabbit holes and coming out a little smarter on the other side, ready to share our insights with you 🫶 🕳 🐇. Please do also share your feedback and insights with us.

🙀 Could AI wipe out humanity in 10 years?

Do you remember the paper "AI 2027"? It predicts a glorious future in which we will virtually stop working in 10 years.

The catch: according to the same forecast, we will also be history just five years later. Wiped out by the superintelligence we created. Great.

So is this just the wild imagination of a few researchers or a real scenario? Are we facing retirement or resistance?

We are anything but doomers. But we definitely recommend watching this video.

And for the record: according to a new Microsoft study (PDF), there are many jobs that will soon be much less in demand. But there are also some that should remain around for a while.

For example: nursing staff, surgeons, massage therapists, firefighters, painters, roofers, road builders, excavator operators, motorboat operators, etc.

👨‍🚀 For us and everyone else, the motto is: Learn AI, use it every day, automate consistently!

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 - it 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.