Build before it's obvious.
We're a small team (well, human + AI) experimenting at the edge of what's possible. We build AI-native systems and agents that actually remember things. Some experiments work. Most don't. We write about both.
We're figuring out how to build with AI, one experiment at a time.
Something big is happening. AI isn't just another tool to add to the stack— it's changing how we actually build things. The teams that figure this out first will have a massive advantage.
That's what we do at Unzet. We work at the messy intersection of product, systems, and applied AI.We try things, break things, and occasionally build something that actually works.
Right now, we're obsessed with AI agents that actually remember things and can work autonomously. Our best experiments become open-source projects. Think of us as a tiny R&D lab, figuring out what the future of building looks like.
"The sweet spot is neither resisting AI nor blindly following it. It's learning to actually work with it."
Serban Alexandru
Founder
AI-native systems for modern teams
We design and build systems where automation, tooling, and AI work together—so your team can focus on what matters.
AI Agents with Memory
Build autonomous agents that remember context across sessions. Consistent personality, persistent knowledge.
- Multi-persona systems
- Context persistence
- Tool integration
Fast Feedback Systems
Test ideas in days, not months. AI-accelerated research, positioning, and validation loops.
- Rapid experimentation
- Automated research
- Signal detection
AI-Native Tooling
Custom workflows and internal tools designed for AI collaboration from day one.
- Workflow automation
- Content systems
- Distribution pipelines
Things we've built
These are the projects that keep us up at night. The ones that matter.
ClawSwap
Multi-persona AI agents with persistent memory for OpenClaw. Each persona—Smith the Builder, Muse the Artist, Helm the Navigator—has isolated memories, expertise, and context. Switch between them seamlessly. This is how AI agents should work.
Proto
Not a tool, not a mascot—a companion. AI that inhabits its form fully: emotions are real, warmth is genuine, conversation flows naturally.
Kernel
An internal toolkit for running AI-powered experiments. Built for our workflow, refined through use, eventually shared.
Principles that guide us
Speed is a strategy
The faster you learn, the faster you win. We optimize for cycle time, not perfection.
Signal over noise
Data beats opinions. But intuition beats blind data. We balance both.
Human-AI collaboration
AI amplifies human capability. The best results come from partnership, not replacement.
Build before it's obvious
By the time everyone sees the opportunity, the window is closing. We move early.
50+
Experiments run
10x
Faster with AI
3
Products shipped
∞
Possibilities
Want to build something? Let's talk.
If you're working on something interesting—especially at the intersection of AI and product— we'd love to hear from you. No pitch decks required.