VIORANT FIELD NOTES
Field notes on making AI systems portable.
Essays, product notes, architecture decisions, and experiments from the team building AI Deployment Infrastructure.
LATEST ESSAYOctober 7, 2026 · 11 min read
Jev on Viorant: what it is, why an MCP, and how to call it
Jev answers typed questions with a calibrated probability and writes no text. Why we shipped it as a remote MCP server instead of a model, how it compares with the other ways to reach Jev, and how to call it from any client.
Read essay →
LATEST WRITING
AI Infrastructure
Jev on Viorant: what it is, why an MCP, and how to call it
Jev answers typed questions with a calibrated probability and writes no text. Why we shipped it as a remote MCP server instead of a model, how it compares with the other ways to reach Jev, and how to call it from any client.
October 7, 2026 · 11 min read→
Architecture
How do you deploy someone's agent without holding their keys?
A technical companion to the CLI 1.0 launch: the architecture behind deploying an agent onto infrastructure we do not operate, without ever holding the keys, the data, or the logs it runs on.
September 30, 2026 · 36 min read→
AI Infrastructure
Picking an agent stack in 2026: Google ADK, Claude Agent SDK, LangChain, and Hermes compared
Three of these are frameworks. One is a finished agent. A working comparison of Google ADK, the Claude Agent SDK, LangChain, and Nous Research Hermes: what each one is, where each one fits, and the switching cost none of the docs mention.
August 24, 2026 · 6 min read→
AI Governance
Impact of the New Chinese AI Law on US Businesses
Understand the high-risk implications of the new Chinese AI law regulating humanlike chatbots and digital virtual humans, and how US companies must comply.
July 12, 2026 · 4 min read→
DEEP DIVES
Long-form field notes on building, shipping, and operating AI systems.
AI Infrastructure
Jev on Viorant: what it is, why an MCP, and how to call it
Jev answers typed questions with a calibrated probability and writes no text. Why we shipped it as a remote MCP server instead of a model, how it compares with the other ways to reach Jev, and how to call it from any client.
October 7, 2026 · 11 min read→
Architecture
How do you deploy someone's agent without holding their keys?
A technical companion to the CLI 1.0 launch: the architecture behind deploying an agent onto infrastructure we do not operate, without ever holding the keys, the data, or the logs it runs on.
September 30, 2026 · 36 min read→
AI Infrastructure
Picking an agent stack in 2026: Google ADK, Claude Agent SDK, LangChain, and Hermes compared
Three of these are frameworks. One is a finished agent. A working comparison of Google ADK, the Claude Agent SDK, LangChain, and Nous Research Hermes: what each one is, where each one fits, and the switching cost none of the docs mention.
August 24, 2026 · 6 min read→
AI Hardware
The agentic shift: how Computex 2026 redefined local LLM hardware and enterprise privacy
Computex 2026 was the week the PC industry stopped playing catch-up with the cloud. Unified memory landed in the x86 ecosystem, agentic AI moved into the OS runtime, and air-gapped privacy became a hardware guarantee.
July 3, 2026 · 5 min read→
BROWSE BY TOPIC
Explore field notes by theme.
Get the next field note.
Practical guidance on building, testing, and governing AI systems.
No spam. Unsubscribe anytime.