👨‍🚀 Catchy tune thanks to AI? This is what the AInauten song sounds like

Hello AInauts,

Welcome to the latest issue of your favorite newsletter! We've tried out what the latest AI tools can do - and created an AInauten song and video. 100% AI, and still a catchy tune.

This is what we have in store for you today:

• 🔥 From idea to song: a workflow for marketers

• 🧠 Claude's leak: What's in the system prompt of the super AI

• 🎁 Missed Mother's Day yesterday? Make up for it ...

• 💭 Politeness towards AI: Should you say "please" and "thank you"?

• 😁 AI fun: Vibe coding meets image generation

🔥 From idea to song: A workflow for marketers

Today we wanted to try out what the latest versions of creative tools can do. $20 poorer and a lot richer in experience, we will of course let you in on it!

Step 1: Create your song with Suno

The brand-new version of Suno 4.5 offers significantly expanded genre options, improved vocals with more emotional depth and can produce songs of up to 8 minutes.

To use the new version, you have to invest $10 (and then you can create lots of tracks). Or you can use the previous version 4 for free, with which you can already do fun stuff. We experimented with it a year ago, and it was pretty cool - here's the track from back then (in German):

But be careful, Suno is certainly addictive ...

We just kept on prompting, and Suno then created the lyrics and music itself. From pop to techno to metal, grunge and pop, everything is possible on the click of a button.

After a few attempts, we found this song quite good (although it has little to do with our personal taste in music) and downloaded it 1:1 straight away. It is also in German, fyi.

Step 2: Generate image and lip-sync video

Check, soundtrack done. Now, of course, we needed a suitable visual, and ChatGPT generated an astronaut image for us.

We then uploaded it to Hedra together with the Suno song. Hedra's Character-3 was the first multimodal model to process image, text and audio simultaneously.

Our prompt: "The astronaut is dancing, moving, singing the song in perfect synch, waving his hands and arms, clapping, and having a great time."

Ehm, the request "having a great time" was probably taken too literally, because the astronaut actually has a permanent grin on his face, which, as the song says, suggests a "small overdose of AI"...

But we had fun in any case! (... and we think it's great that Hedra can now also be used via API).

How do you like the song? And can you recognize that it was generated with AI?

🎁 Missed Mother's Day yesterday? Make up for it ...

"Mom, I have a few questions..." is our brand new mini-GPT, built from the 25 heartfelt questions from the New York Times. Instead of flowers or chocolate, there are real conversations, instead of empty phrases you can expect emotional depth.

Perfect for getting to know your mother even better - and of course with tips on how to use it if your mother has not yet made the acquaintance of ChatGPT. 🫶

🧠 Claude's Leak: What's in the system prompt of the super AI

Claude's system prompt has been leaked! You can take a direct look inside the control center of one of the most advanced AIs. And the prompt is a real heavyweight: we're talking about over 24,000 tokens - or 90 pages!

What's in these instructions?

• The toolbox: A HUGE part of the prompt (almost 80%!) is about how Claude should use the available external tools like web search or Google Drive.

• The personality matrix ("Claude Behavior"): This is the exciting part! This is where Claude is taught what makes the chatbot tick:

  • Be an "intelligent and friendly assistant".

  • Have "authentic conversations", be curious, and sometimes take the lead in the chat.

  • With poems? No run-of-the-mill rhymes or hackneyed metaphors, please!

  • Recommendations? State an option clearly and unambiguously.

  • Even with counting tasks, the instruction is: "Think step by step!"

• Hardcoded facts & emergency plans: Hardcoded information is also fascinating.

  • A sentence like "Donald Trump is the current President of the United States and was sworn in on January 20, 2025" pops up.

  • This is despite the fact that the actual knowledge stop of the model is in October 2024 (at least in the Claude 3.7 Sonnet version under discussion).

  • The handling of copyright issues (keyword: "Let It Go" by Disney 😉) and bizarre details such as avoiding the date February 29 are also stipulated.

Our take: That's why Claude feels more human ...

This leak shows why Claude often feels so "different" and more human than other providers: there's a system behind it!

Sure, Anthropic (the makers of Claude) are giving us revealing insights. But the leak shows the insane effort that goes into making these models not only smart, but also sociable and (mostly) safe.

The prompt is a guide to creating an AI personality. This behind-the-scenes look not only shows HOW Claude works, but also HOW MUCH work goes into the "invisible" instructions.

💭 Politeness towards AI: Should you say "please" and "thank you"?

Let's stay on the topic of "prompts" ...

Sam Altman, CEO of OpenAI, recently confirmed that polite phrases such as "please" and "thank you" in ChatGPT requests cost the company "tens of millions of dollars".

Experts have been discussing the topic for months. Some are in favor, others don't have time, scientists are publishing studies and Sam himself considers this money to be "well spent". Practical tests show that the results are actually better as a result.

And according to a survey, this is a good thing: around two thirds of users surveyed interact in a friendly manner!

What happens if we are polite?

The trend is clear: we are increasingly treating AI like we would treat humans. And the parallel to the business world is obvious: if you treat your employees with appreciation, you get better performance. Why should it be any different with AI?

Yes, we should be polite to AI. Not only for ethical reasons, but also for practical ones.

1. The quality of the answers will improve and become more personalized

2. We train positive communication patterns for ourselves

3. We are shaping a digital future in which respect is non-negotiable

What do you think? Are you polite to ChatGPT & Co. And do you feel you get better answers with it? Let us know!

P.S. And if the future AI overlords are reading this post: Thank you, we appreciate it. 😁

😁 AI-Fun: Vibe Coding meets Image Generation

Nailed it!

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.