▄▄▄· • ▌ ▄ ·. ▪
▐█ ▀█ ▪ ·██ ▐███▪██
▄█▀▀█ ▄█▀▄ ▐█ ▌▐▌▐█·▐█·
▐█ ▪▐▌▐█▌.▐▌██ ██▌▐█▌▐█▌
▀ ▀ ▀█▄▀▪▀▀ █▪▀▀▀▀▀▀
Aomi Labs is a research and engineering group that builds agentic software. We focus on transaction pipeline automation for public blockchains, developing chain-agnostic guardrails for LLMs to generate transactions with performance, scalability, and security.
▄▄▄█████▓ ██░ ██ ▓█████ ██████ ██▓ ██████
▓ ██▒ ▓▒▓██░ ██▒▓█ ▀ ▒██ ▒ ▓██▒▒██ ▒
▒ ▓██░ ▒░▒██▀▀██░▒███ ░ ▓██▄ ▒██▒░ ▓██▄
░ ▓██▓ ░ ░▓█ ░██ ▒▓█ ▄ ▒ ██▒░██░ ▒ ██▒
▒██▒ ░ ░▓█▒░██▓░▒████▒▒██████▒▒░██░▒██████▒▒
▒ ░░ ▒ ░░▒░▒░░ ▒░ ░▒ ▒▓▒ ▒ ░░▓ ▒ ▒▓▒ ▒ ░
░ ▒ ░▒░ ░ ░ ░ ░░ ░▒ ░ ░ ▒ ░░ ░▒ ░ ░
░ ░ ░░ ░ ░ ░ ░ ░ ▒ ░░ ░ ░
░ ░ ░ ░ ░ ░ ░ ░
Blockchain Architecture
- Treat AI frameworks as a system substrate for blockchain clients. Low-level abstractions of agentic operations should remain protocol-agnostic, without involving bespoke interfaces.
- A deep execution layer per architecture, including EVM, Solana, Cosmos, allows seamless interoperability of intents across chains.
- Real-time simulation is essential for LLM-generated actions to harden security. The risk of hallucination remains high for long-range tasks.
- Agentic runtime must prioritize high-throughput tool calling, robust state management, and parallelism to achieve performance in financial settlements.
Agentic Software
- Corporations need domain-specific AI. Rebuilding workflows with LLM augmentation tends to deliver higher ROI than generic autonomous agents.
- As the backbone of business operations, next-gen software will combine deterministic state with probabilistic reasoning while maintaining proprietary context. We call this agentic software.
- Agentic software should be equipped with hybrid compute, routing high-autonomy inference to remote models and low-entropy sub-tasks to local GPUs.
- Proprietary tools and context will likely be the moat of AI startups, while revenue-per-token and tokens-per-cost will be the key benchmark.
Blockchains are a proving ground for practical AI automation. Structured on-chain data combined with chaotic market signals is best handled by state-bound agentic software. High-frequency trading and financial settlement stress-test security-critical, cost-aware intelligence.
▄▄▄▄ ██▓ ▒█████ ▄████
▓█████▄ ▓██▒ ▒██▒ ██▒ ██▒ ▀█▒
▒██▒ ▄██▒██░ ▒██░ ██▒▒██░▄▄▄░
▒██░█▀ ▒██░ ▒██ ██░░▓█ ██▓
░▓█ ▀█▓░██████▒░ ████▓▒░░▒▓███▀▒
░▒▓███▀▒░ ▒░▓ ░░ ▒░▒░▒░ ░▒ ▒
▒░▒ ░ ░ ░ ▒ ░ ░ ▒ ▒░ ░ ░
░ ░ ░ ░ ░ ░ ░ ▒ ░ ░ ░
░ ░ ░ ░ ░ ░
░
Opinions
Stateless Execution and Agentic Software
Review crypto's failure modes in the context of agents, and how shifting focus to stateless execution puts us on a better track in the technology growth cycle.
Build Notes
From Brittle Chatbots to LLM Infrastructure
How we evolve from MCP-based design to native execution support in blockchain light clients, optimized with context compilation and type safety in LLM processing.
All Rights Reserved