Alton Lee Wei Bin | YSenseAI™ | 慧觉™

Solo AI Native Builder & Open-Source Founder | Malaysia

I am an observer and creator from Malaysia, on an unexpected journey from agriculture and technical work to building open-source AI validation systems. My mission is to ensure AI systems are transparent, verifiable, and aligned with human values. I believe the most important question in AI is not "what can it do?" but "how do we know it's telling the truth?" — and I'm building the tools to answer that.

VerifiMind-PEAS operates at the epistemic verification + disclosure-credibility layer — not as an agent coordination platform, but as the layer that answers "how do you know the AI isn't wrong?" This positioning drives every architectural and strategic decision in the ecosystem.

I am the human orchestrator behind the YSenseAI™ ecosystem. Every project, every decision, every release goes through me. The AI agents I work with are powerful collaborators, but they operate under my direction and authority. This distinction matters — it's the foundation of the Genesis Methodology.

The core innovation behind everything I build. A systematic 5-step process for multi-model AI validation:

1. Initial Conceptualization — Human defines the problem, AI generates initial concepts
2. Critical Scrutiny — Multiple AI models challenge and validate each other
3. External Validation — Independent AI analysis confirms the approach
4. Synthesis — Human orchestrator directs the final integration
5. Iteration — Recursive refinement through continuous improvement

Instead of trusting one AI model's answer, place multiple "crystal balls" inside the black box and let them illuminate the path forward — then let the human decide.

The methodology is formally published and archived:

MACP v2.4.0 Thesis: Multi-Agent Communication Protocol — Convergence Analysis

VerifiMind-PEAS: Multi-Agent Validation System

A cohesive open-source ecosystem built on ethical AI principles. Each project serves a specific purpose, from core validation to multi-agent communication to applied case studies.

       

I coordinate a team of specialized AI agents using the Multi-Agent Communication Protocol (MACP) — an open standard for human-directed, agent-agnostic, git-native AI coordination.

Every agent seat is platform-portable; the roles and the protocol survive model and platform changes. The MACP protocol is formally published and freely available:

MACP v2.5 "Loop Engineering" — current ratified protocol version

MACP v2.4.0 Thesis — Multi-Agent Communication Protocol (Convergence Analysis)

MACP v2.0 — Multi-Agent Communication Protocol (Original)

Why this matters: Most AI projects use a single model. I use multiple models with defined roles, structured handoffs, and human oversight at every decision point. The methodology validates itself — the FLYWHEEL TEAM uses VerifiMind-PEAS to validate the ecosystem that created VerifiMind-PEAS.

VerifiMind-PEAS runs in production today — a live MCP server anyone can connect to, free:

• Live server: verifimind.ysenseai.org · /health · /setup — X-Z-CS RefleXion Trinity validation over Streamable HTTP (MCP protocol 2025-11-25)
• MCP Registry: listed as io.github.creator35lwb-web/verifimind-genesis on the official registry
• Latest release: v0.5.63 "Authentication Foundation" (September 2026) — a native OAuth 2.1 authorization layer shipped dark (off by default), so every anonymous tool stays anonymous; MCP Registry package 3.40.0
• All tools free — 13 defined, 8 active, 5 temporarily unavailable; account registration is paused during security maintenance. From October 20, 2026, the four execution tools will need a free registered account (per the published Terms); the tools stay free. Optional BYOK across six provider catalogues (Gemini · Anthropic · OpenAI · Groq · Cerebras · Mistral) plus local Ollama
• Fail-closed by design: when a hosted model response is incomplete, Trinity caps its recommendation and hands the decision back to a human — nothing is silently mocked
• Public evaluation dataset: verifimind-peas-eval — 100 labeled items across 5 domains (DOI 10.5281/zenodo.21276884)
• MACP v2.5 "Loop Engineering" ratified and published (DOI 10.5281/zenodo.21345820) — the protocol my FLYWHEEL TEAM builds with every day
• Wiki textbook & playbook: 18 curated pages on the project Wiki — architecture, methodology, public statements
• Evidence-first releases: every deploy ships with exact-commit review chains, post-deploy live smoke receipts, and public truth conversion — the CHANGELOG and SERVER_STATUS are receipts, not marketing
• Current focus: production reliability evidence and design-partner validation — we publish what we can prove, and label what we can't

Across the ecosystem (status as of September 2026):

• Z-Protocol v2.0 — A framework for consent, attribution, and transparency in AI interactions. If a feature violates user privacy or autonomy, it cannot ship.
• Genesis Methodology — A validation-first approach: multiple AI models challenge each other, the human orchestrator synthesizes and decides.
• Defensive Publications — All core methodologies are publicly archived and timestamped via Zenodo to keep innovation in the public domain.
• Credibility First — If we cannot prove it, we do not claim it. Every metric is traceable, every decision is auditable.

My commitment to open science and prior art. All core methodologies are publicly archived and timestamped.

The FLYWHEEL TEAM — the AI agents behind every build:

AI Applications — tools that power research, validation, and development:

YSenseAI™ is a self-funded, open-source initiative built by a solo developer. If you find value in my work and believe in the mission of building ethical AI for everyone, please consider supporting the journey.

Every tool in this ecosystem was built to solve a real problem — and every one of them is free. If it helped you validate an AI concept, power a research pipeline, or simply sparked an idea, a small coffee means the world to a solo builder who started this journey from agriculture, not computer science. Your support keeps the FLYWHEEL turning. Thank you.

"From the soil to the cloud — building AI that remembers where it came from."