Building a "Sovereign AI Farm" in Sweden – Launching tomorrow with 0 followers. Thoughts?
I’m launching agentfarm.se on Product Hunt tomorrow.
Beyond building a hub for sovereign and secure AI agents, this project is my way of diving deep into how agents actually behave and how we can influence them. I’m a firm believer in the idea that the best way to understand agentic workflows is to build them in the open.
The Goal:
To create a platform where we can learn together. I want to use the feedback from the first users to steer the development of the farm. Instead of building in a silo, I want to see how others interact with these agents and what security/sovereignty trade-offs they are actually concerned about.
The Strategy:
I'm a solo maker launching without a large following. I’m treating this launch as a "live lab" – a way to gather insights and iterate fast based on real-world usage.
I’d love your thoughts on:
What’s the biggest "black box" mystery you’ve encountered when working with AI agents?
As a community, how can we better influence agent behavior without relying on proprietary guardrails?
The Demo:
You can check out the pre-launch and the video here: https://www.producthunt.com/products/agentfarm/agentfarm/prelaunch
I’ll be here to discuss agent mechanics, the tech stack, or how we can build better local AI together.
The sovereignty angle is smart positioning. With all the data residency regulations tightening across the EU, having compute that explicitly stays within Swedish jurisdiction is a genuine selling point — not just a buzzword.
Launching with 0 followers is honestly fine. Most successful PH launches I've seen came from founders who spent the 2 weeks before launch engaging authentically in communities (exactly like posting here). The followers come from the conversations, not the other way around.
Curious about the business model though — are you going usage-based or flat subscription? For AI compute especially, usage-based seems like the natural fit but it makes revenue prediction harder as a bootstrapper.
Also: what's your approach to monitoring usage across the different AI providers your agents connect to? That's become one of the biggest pain points I've seen — people spinning up agents across Claude, GPT, Gemini etc. and having no idea how close they are to hitting rate limits until they actually hit them.