Status: Experimental Research & Neural Prototyping (Alpha) 🧪
Exploring systemic security through minimalist intelligent agents.
The Synapse module is an experimental laboratory within the AOXCON ecosystem. Our primary objective is not to build monolithic AI, but to deploy minimalist, high-integrity agents designed to enhance blockchain security and coordination.
We believe that small, specialized "neurons" (agents) are more resilient, verifiable, and secure than complex, opaque systems.
- 🛡️ Micro-Agent Security: Use lightweight agents to monitor cross-chain state transitions and detect anomalies in real time.
- 🧠 Decentralized Logic: Research small-scale neural patterns that can support decision-making across XLayer, Sui, and Cardano.
- 🔐 Verifiable Integrity: Prioritize transparency and simplicity so each agent action remains predictable and auditable.
- Conceptual framework for micro-agents
- [/] Prototyping "Neuron-0" security monitoring flow
- Multi-chain state synchronization tests
The repository includes a fully isolated training campus for building domain-specific mini AOXCAN assistants.
- CLI Generalist Assistant (foundation)
- XLayer Operations Specialist
- Sui/Move Specialist
- Cardano Integration Specialist
- Cross-Chain Mainnet Guard (specialist coordination)
Each track gets independent artifacts so experiments never collide:
datasets/docs_en/curriculum/algorithms/checkpoints/eval/logs/reports/exports/
python -m academy.cli --base-dir . bootstrap
python -m academy.cli --base-dir . list-tracks
python -m academy.cli --base-dir . validate
python -m academy.cli recommendLegacy bootstrap script is still available:
python scripts/bootstrap_training_campus.pyThe command set generates and validates training/campus/manifest.json with track domains, algorithm plans, and readiness gates.
Short answer: yes, if you keep strict promotion gates.
- Keep domain-specialized tracks (XLayer / Sui / Cardano / CLI) isolated.
- Use L1 as a mandatory base before promoting to L2/L3 specialists.
- Treat
validateas a quality gate in CI so broken tracks cannot progress. - Prefer iterative open-source contribution per track instead of one huge training run.
- CLI Generalist Assistant (foundation)
- XLayer Operations Specialist
- Sui/Move Specialist
- Cardano Integration Specialist
- Cross-Chain Mainnet Guard (specialist coordination)
Each track gets independent artifacts so experiments never collide:
datasets/docs_en/curriculum/algorithms/checkpoints/eval/logs/reports/exports/
python -m academy.cli --base-dir . bootstrap
python -m academy.cli --base-dir . list-tracks
python -m academy.cli --base-dir . validateLegacy bootstrap script is still available:
python scripts/bootstrap_training_campus.pyThe command set generates and validates training/campus/manifest.json with track domains, algorithm plans, and readiness gates.
This project is intentionally open and iterative. If a training approach fails, the structure is designed so others can continue the work cleanly, track by track, with reproducible outputs. Built with integrity for a decentralized future.
Bu repo artık sadece klasör açan bir taslak değil; mainnet'e yakın, görev uzmanı mini AOXCAN yetiştirmek için seviyeli ve izole bir eğitim sistemi sunar.
- XLayer Ops Assistant
- Sui/Move Assistant
- Cardano Integration Assistant
- CLI Generalist Assistant
- Cross-Chain Mainnet Guard (uzmanları koordine eden üst seviye)
- Her track tamamen izole (
datasets,docs_en,curriculum,algorithms,checkpoints,eval,logs,reports,exports) - İngilizce doküman kaynakları için ayrı
docs_en/alanı - Birden fazla algoritma ailesi ile karşılaştırmalı eğitim (LoRA, QLoRA, DPO, MoE, vb.)
- Track bazlı mainnet readiness gate tanımı
Bu depo artık farklı seviyelerde, birbirine karışmadan mini-agent eğitimleri yürütmek için izole track yapısını destekler.
- L1 / from-scratch: Sıfırdan eğitim denemeleri
- L2 / pretrained-adapter: Hazır modeller üzerinde güvenli ince ayar
- L3 / hybrid specialist: Göreve özel hibrit uzmanlaşma
Her track için datasets/, checkpoints/, logs/, reports/, exports/ klasörleri ayrı açılır. Böylece sonuçlar birbiriyle çakışmaz.
Başlatmak için:
python scripts/bootstrap_training_campus.pyBu komut training/campus/manifest.json içinde tüm track'leri, domain kapsamını, algoritmaları ve readiness gate bilgilerini üretir.
Bu komut training/campus/manifest.json üretir ve eğitim alanlarını otomatik kurar.