AI is the hottest shit right now and as everyone else I can’t avoid this topic. As a semi tech and tech interested person I always want to try the new things out hands on myself and make my own judgment instead of listening to either overly positive or as well negativ statements by popular voices.
AI me generated with flux/LoRa
This is the place where I try to summarize my experiences with different use cases of AI. Mainly in the field of my business as marketing person but also other related fields.
Vibe Coding
I redesigned https://anaidea.com/ completely using the new gemini 3 pro model which can create nice looking interface that don’t look like AI generated (Dez 2025)
One of the first vibe coded apps was a Google notes clone I created with bolt.new beginn of 2025. I steady worked on it and release in summer 2025 fleeknote.com. In November/Dezember I re-factored large parts of the app. It was clearly visible in the code that the early 2025 vibe coding tools were not as good. There was a lot of duplicate code, anti-patterns and workarounds used – in short: ugly code base. (Nov/Dez 2025)
A json prompt generator for using with Veo3 or other video generating AI models: https://veo3json.moweco.com/ (Aug 2025)
Created a simple AI business idea validator using bolt.new and cursor: https://anaidea.com/ (July 2025)
I tried again (May 2025) some popular vibe coding tools to create a multi agent research system. This time we have a clear winner.
My recap on coding again after a long break with the help of AI: AI software development – recap from a non dev person
My take on creating apps with online generators like lovalble, bolt, v0 or replit: The sad reality of magical AI powered dev environments (Jan 2025)
For my consultancy I created a marketing & business trends agent (RAG): https://v0-assistant-api-chatter.vercel.app/
Simple tools and forms I created with AI (with Claude):
Vibe game development
A simple starfleet commander clone proof of concept with three.js done with lovable: https://starfleet-web-simulator.lovable.app (Nov 2025)
Some try outs and smaller pieces:
Pseudo3d projection tryout (Aug 2025)
A clone of the famous paddle war inside of Comander Keen 4 (Aug 2025)
Simple top down soccer game prototype (Sept 2025)
I created a 2d game completly with the help of AI – so code, images, music etc.: I built a video game – with the help of AI (Jan – Jun 2025)
Creating a complete game sprite in chatgpt (gpt-image-1 model):
The result is good and looks nice on first sight but when you look closer it is not usable for an actual game.
AI Image Generation
Peak performance model comparison (Dez 2025)
There is a lot going on with image generation models and I have compared the 5 currently best model with a simple prompt to create a vintage style realistic photo:
A photorealistic realistic, candid medium shot of a young woman with a radiant, wide smile and long, dark, wet hair. She is wearing a blue ribbed crop top and layered silver necklaces. Her skin is glowing with a slight sheen of water/sweat. The setting is a beach club at dusk with a vintage film aesthetic (35mm grain). The background is softly blurred, showing orange beach chairs, a thatched roof, and a sandy beach under a hazy, overcast sky. Muted, warm tones with a low-contrast, nostalgic feel.
For me the flux2pro model is winning here very close in front of the new 1.5 gpt model from openai. Both really can generate the look & feel with naturally looking faces. Also you can in this test clearly see the bias of training data as the Qwen model has create an Asian woman, although no ethnicity was specified in the prompt.
Professional Photos generated with AI compared to 1 year ago (Oct 2025)
The 3 best AI image models head to head comparison (Oct 2025)
The nano banana hype is real (August 2025)
created on lm arena this pic with a simple prompt: put the cast of the original star trek series into this star wars movie poster
New edit functions in gpt-image and flux-kontext (April 2025)
Training AI models and generate custom portrait shots (Okt 2024)
I wrote a couple of articles outlining the process of training a model with your own face to create portrait shots of you:
Comparing different models (April 2025)
This is a small excerpt of my test with different image generating models. flux is currently really fabulous when it comes to realistic images but has its flaws when creating more graphical visuals (see last example below). imagefx from Google seem to be also a very well choice. Disappointing was my test of Janus-pro which still has huge problems with faces.
flux 1.1 pro vs gpt-image
Dall-E 3 vs Janus-pro
imagefx from Google and Dall-E 3 comparison
flux 1.0 vs. Dall-E 3
flux 1.0 vs. Dall-E 3
Those where for example my first steps: My AI Image generator tryouts
And some newer AI Images including prompts: Some recent Image AI Prompt Highlights