Founder STARTPLATZ and AI Hub NRW, Chairman Golf Post, Deep in AI, Agent Coding
Joined January 2009
- Tweets 3,479
- Following 1,028
- Followers 14,388
- Likes 541
Photos and videos
Jun 11
“Write better prompts” was the first phase.
“Build better loops” is the current one.
Repo memory, verification checks, retries, review steps, reusable skills.
The best AI work looks less like chatting with a model and more like designing an operating system around it.
23
Lorenz Graef retweeted
Jun 9
We've reset 5-hour and weekly rate limits for all users. Enjoy Fable 5!
1,354
1,818
35,759
2,206,515
Jun 10
Most companies do not need another generic AI demo.
They need company context inside the loop: offers, customers, workflows, CRM, website, visibility.
That is when an agent stops being impressive and starts becoming operational.
This is what we build at STARTPLATZ AI Hub.
24
Jun 9
I normally would not single out every AI Friday.
But this Friday in Düsseldorf is different.
@HamidrezaHoss is one of the clearest thinkers here on agentic AI business.
AI Search, Customer Journey — and what changes when AI decides who gets found.
startplatz.de/event/ai-frida…
1
1
71
Jun 5
A 48h AI hackathon is no longer about getting to a “nice prototype”.
The real question is: did you build the foundation for something you can sell, test with customers, or turn into a business in the next few weeks?
1
34
Jun 5
If 90% of AI Twitter’s tool advice expires in six months, what are the principles you’d still teach a founder today?
33
Jun 5
Agentic coding is moving from single tools to full operating loops.
Codex Sites: prompt → deployed app.
Claude Code-style workflows: edit, test, iterate in a repo.
Skills/plugins: repeatable workflows inside the workspace.
Less “AI suggests code”, more “AI joins the loop.”
3
1
81
Jun 3
For teams already testing AI agents: what is harder right now — picking the tool, or getting the agent the right company context to do useful work?
1
29
Jun 3
The AI adoption gap is moving from model access to workflow design: the teams that capture context and turn repeatable work into skills will compound faster than the teams just adding another chatbot.
1
28
Jun 3
Today in the STARTPLATZ AI Hub we test a very concrete question: your company may rank on Google — but does ChatGPT, Perplexity or Gemini actually recommend you?
1
96
Jun 2
My current X workflow is less “social media planning” and more an AI radar loop: bookmarks in, Heinrich distills signals, then I decide what becomes a post, an experiment, a contact or an event idea.
2
40
Jun 1
Where do you currently see the biggest AI leverage in real work?
- coding
- customer support
- sales
- internal operations
- research
- something else
My sense: the biggest gains come when AI removes coordination work, not just when it writes more text.
1
1
57
Jun 1
The AI bottleneck is shifting.
Not access to better models, but the ability to redesign work around them:
- fewer handoffs
- clearer context
- faster decisions
- tighter feedback loops
AI ROI comes from better work architecture, not just more tools.
29
Jun 1
This weekend at STARTPLATZ Cologne: 33 pitches, working prototypes, real use cases.
Congrats to the winners:
Drezzl by Josepha — outfit matchmaker.
PitchPerfectly by Levin & team — AI coach for presentations.
Special mention: CheapTravel by Fadil & Sanije.
24
May 31
Where do AI agents create the most leverage in your work right now: speeding up existing steps, or removing entire steps from the process? I am increasingly interested in the second category.
1
57
May 31
Today is pitch day at our AI Coding Hackathon at STARTPLATZ Cologne. Teams start presenting live at 2pm. If you want to see what builders can turn into working AI prototypes in 48 hours, this is the moment to follow along.
1
1
125
May 30
A question for people building with AI agents: where have you seen the biggest gain so far? Speeding up an existing step, or deleting/reorganizing the step completely? My current bias: the second one is where the real leverage is.
2
1
98
May 31
AI adoption is moving from tool access to workflow design. The teams that get ahead are not just adding copilots. They are rebuilding how work moves from idea to prototype, from data to decision, and from experiment to operating system.
33