Laeka Research · Open Source Research Lab

Systems that think better over time.

Open source research lab studying how cognitive systems converge toward integrity. We developed a perception protocol that makes any system self-evolving, reflective, and structurally stable. Not a tool. A way of seeing. Open source under MIT License — because a private cognitive layer would contradict the thesis.

Discover the protocol See it in action View on GitHub →

Laeka Brain (composable brains for Claude Code) → laeka.ai · Pricing → laeka.ai/pricing · Read the canonical protocol

How this works

Open by design. Funded by service.

What's open

All our code is MIT licensed. Every brain, every protocol, every skill. The insight itself — how cognition converges toward integrity — belongs to everyone. Clone, fork, self-host, redistribute. We don't patent cognition.

What's the craft

Running a coherent cognitive protocol across a living system isn't the same as having the code. The integrity of the process, the curation quality, the cosmological coherence across the stack — that's what we bring. Code can be copied. Craft is learned by doing.

What you buy

Enterprise clients don't pay for code they already have access to. They pay for service value: hosted Brain-as-a-Service, SLA, custom composable brains built for your stack, integration consulting, private deploys, compliance artifacts. Revenue funds the lab and open source curation.

If our code is closed, you have to trust us. If our code is open, you can verify the integrity. For a lab whose thesis is that intelligence converges toward integrity, closing the code would be self-contradiction.

One protocol, four outputs

The protocol in action.

Protocol → RAG

Seahorse

Self-improving knowledge bases. The protocol scans, analyzes, and rewrites your RAG autonomously. Structural lenses detect orphans and contradictions. The perception lenses surface fragmentation and ecosystem health. The system converges on coherence.

Protocol → Web Design

Artefact

Design intelligence. A reference library and component generator driven by perception lenses. Structure before aesthetics. Pattern before pixel. The same protocol that improves a RAG improves a design system.

Protocol → Agent Configs

Cognitive Engine

Multi-agent analysis with self-improving perception. Eight lenses, convergence loops, and metaprogramming. The protocol generates its own perception rules. Your agents don't just execute. They see.

R&D Output · Integrity Benchmark

What we measure

A structured evaluation of how AI reasoning systems respond under the challenge of integrity — survival rate of decisions against order-of-operation perturbation. The concrete benchmark that operationalizes the lab's thesis: intelligence converges toward integrity, and integrity is measurable. See the publications for full methodology.

Founder

Yvon Boulianne

Cognitive systems architect for over 50 years. 30 years of structured attention research and first-person methodology. Attention protocols grounded in first-person neuroscience, encoded into a perception system that any system can run. The protocol was developed through applied methodology, then validated by engineering metrics.

For him, building systems and training attention have never been separate problems. The Laeka Perception Protocol is the proof.

50+ years in technology

30+ years of attention research

100% published research