The current situation
It has now been 1 1/2 years since I started this project. In my perception, the world is spinning faster and faster. The daily news about AI, new models, but also new drone and weapons technologies that fly through space and reach speeds of up to 12,000 km/h, does not stop. On the one hand technically very impressive, but on the other hand also very frightening when you realize that such things are being used not far from here to kill people and cause pure destruction.
The constant news about better LLMs and agents increasingly triggers in me the urge to try out and test the technical innovations. To test what is already possible. Even though I feel more and more behind, because someone just can’t test everything at once.
Time is Running Faster
I notice more and more how my daily work routine is changing. More tasks in less time, because it’s possible with AI. The only problem: who controls all of this? As a human being, you feel left behind, almost stupid, because you just can’t keep up with what’s possible with AI anymore. The work that dozens of developers used to do in many days of work is now handled by significantly fewer developers with AI, or simply taken over completely by agents.
I feel more drained after work and increasingly unable to spend even more time in front of the screen. I am torn between the ambitious goal of wanting to build something of my own, but at the same time not wanting to spend more time at the screen, as the actual world without a screen falls short and I think there is so much more that one should enjoy.
Nevertheless, I have still put a lot of time into the development of my app, even though no one else has seen or tested the app yet. I keep thinking there are still too many bugs in it and not everything really works yet.
In the meantime, I was closer to the idea of founding my own company to bring the app to the App Stores. I had meetings with startup consultancies and presented my pitch deck. While these talks resulted in many tips and contacts, they didn’t really give me the courage to actually go through with a foundation although I was very pleased to have someone that listens to my ideas. The opinion of the people I met was that this what I built was quite impressive. Of course, that makes me happy.
Pitchdeck Slides:
Doubts
The doubts that what I’ve built isn’t good enough and that no one will use it anyway are huge for me. At least as big are the legal doubts and the doubts about whether everything the AI has built is good enough and secure. In the meantime, I have moved away from a cloud-centric approach and had the idea of running the app exclusively on the smartphone. On-device / Edge LLMs like Google’s Gemma E2B or a Liquids LFM are promising. I think this will be a strongly developed field in the future by Apple and Google, even if there will always be a cloud connection since the big corporations depend on the data.
Porting the backend from Python to Rust was no problem for Claude Code in a few hours. I believe this would be a viable path for me to avoid security and data protection responsibilities and create a solution that runs exclusively on the user’s device. The disadvantage that arises from this is the further development of the project. Over time, I’ve noticed that LLM and agent evaluations are very time-consuming and rely on user data. One could enable the user to share data for further development. I think this is a good way to give the user the choice of whether they want to share their data for further development or not.
