I ve spent the last few months obsessed with a specific failure in multi-agent systems: The Black Box Dilemma.
Most MAS setups I see (including my own v1) treat orchestration as "glue code." You have a central manager, a few agents, and a lot of hidden logic. When it fails, you re left guessing. As someone on IndieHackers put it: "You re just parallelizing chaos."
The Shift: From "Dictator" to "Blackboard"
In our latest iteration, we deleted the 430-line monolithic controller. Instead, we implemented a Blackboard Pattern.
Hey everyone, been a Product Hunt visitor for years but never signed up, so I thought I'd start by sharing some recent learnings that hopefully other builders will find useful.
For context, we're building FanBase Copilot, an AI assistant for content creators that learns their voice and context over time. The memory layer is critical. It's what makes the AI actually useful after the first conversation.
So I've got a bunch of agents building Klatch for me now, following Gall's law principles. It's still in alpha. No overpromising. But I got it going this past Saturday and it's already functional. Like most of my stuff this is fully open source so anyone can grab it and fork it, or make a version for Gemini, or whatever. Let me know if you find it useful. We're working on context and resource preservation post-import, next.
I didn t build our internal compliance tracking tool the traditional way.
I vibecoded it.
Instead of long PRDs, heavy sprint planning, and weeks of back-and-forth, I stayed close to the problem and built in tight feedback loops shipping small, observing behavior, and iterating fast.
This was a deliberate experiment inspired by my CTO. I wanted to test a simple question: Can a Product Manager ship a real website end-to-end today without handoffs?
