// AI_PROTOCOL

Agentic systems that multiply you.

YambaCode designs autonomous AI architectures that automate the repeatable and free your capacity for genuinely creative work.

AGENTIC AI LLM SYSTEMS GENERATIVE AI AUTOMATION

// CAPABILITIES

Agent architectures

Multi-agent systems that plan, use tools, and act — reliably, with guardrails.

Generative integration

LLMs woven into real products and pipelines, not demos — with evaluation and observability.

Deep-tech strategy

Where rigorous mathematics meets adaptive AI. Outside the box? What box.

// RESEARCH · MATHEMATICS & VERIFICATION

Research you can inspect.

AI-assisted mathematical research gains substance when its arguments can be checked. From the Yamba AI Math lab: an established combinatorial theorem, formalized in Lean and accompanied by its full technical article.

Conceptual mathematical artwork: cyan and amber tiles dissolve into particles, leaving an ordered constellation.

LEAN-CHECKED MATHEMATICS ·

From a Failed Corner Slide to a Verified Tableau Injection

Christopher Yamba’s 11-page technical article develops the corrected tableau construction and completed Lean formalization of Cho–Huh’s established height-two theorem. Its expansion holds for every finite palette of colors. The proof repository contains the exact statements and source-bound verification record.

Author-published exposition and formalization of established mathematics. The paper has not undergone external journal peer review.

Read the illustrated story Inspect the Lean proof and sources

// YAMBACODE LAB · NOW BUILDING

A lab where you can watch AI think.

YambaCode Brain is my private, local-first AI laboratory. It holds a knowledge graph of AI research, answers questions as moving diagrams, runs agents drawn as creatures, and lets you train small networks and look inside them. These silent clips are from my first recorded session with it.

Atlas

717 papers, concepts, theories and repositories as a sky of stars, held in place by their relations.

Abisa Twi for asking

Ask a question and the answer is drawn step by step through the knowledge graph, with a live simulation.

Forge

Assemble a network block by block and watch it learn. Every tile is one neuron, drawn as the function it computes.

Hive

Agents as creatures. Each personality trait reshapes the body, and the same traits steer how the agent works.

What I am working on now

  • A self-improvement loop: a spider-diagram benchmark scores eight capabilities of the lab, and the lab works on its weakest one, keeping a change only when it wins on held-out questions.
  • Growing the knowledge graph from the papers I read each week, so the lab knows what I know.
  • Voice-first answers: ask out loud and hear the explanation while it is being drawn.

See every deck of the lab

// FAQ

Questions, answered.

What kind of AI does YambaCode build?
YambaCode builds applied and agentic AI systems: autonomous architectures, LLM-driven agents, and generative AI integrations that automate real workflows and multiply human capacity.
Does YambaCode do AI consulting?
Yes. YambaCode AB provides AI architecture, engineering and deep-tech strategy. Learn more at www.yambacode.com.