Field notes from building agentic infrastructure.
Technical deep dives, market analysis, research, announcements.
What we are working on, what we have learned, and what we are watching in the broader autonomous-agent space. Posts are categorised by Technical Deep Dives, Market Analysis, Research, and Announcements. Subscribe to get them in email instead of finding them through a search.
THE AI PILOT WORKED. TWO YEARS LATER IT IS STILL A PILOT
Almost every health system has an AI success story that never left the department it started in. The results were good enough to present at board level... and then nothing happened.
AGENTIC AI IS SCALING FAST ACROSS HEALTHCARE. THE BARRIER TO PRODUCTION IS CONTROL
Healthcare is among the fastest moving and the highest stakes for Agentic AI. A swarm of vendors is selling an agent that just handles the work: booking, messaging, triage, follow-up, around the clock. The pitch is compelling. The problem is that this work touches data, records, and patients.
INTRODUCING UNICITY AOS: A MODULAR OPERATING SYSTEM FOR AI AGENTS
You already run AI agents. You just can't trust them. That's the problem we built Unicity AOS to solve, and today the open source Community edition is public.
HOW TO REALISE THE AGENTIC OPPORTUNITY IN HEALTHCARE
Agentic AI is scaling fast across every vertical. In healthcare, the barrier to production is control. Unicity can help cross the chasm
AI AGENT NATIVE SECURITY ARCHITECTURE NEEDS TO LIVE IN THE KERNEL
Most enterprise AI security today operates at the wrong layer. It monitors model outputs, flags prompt patterns, filters responses.
BUILD THE MACHINE ECONOMY WITH UNICITY
Our incentivised testnet has been live for 8 weeks, and as of last week, our Testnet v2 is live, and so is the Sphere Developer Portal. Starting today we are running a four-week call to build the first real apps and agents on the network
UNICITY LABS AND QUANT PARTNER TO BRING VERIFIABLE EXECUTION TO AGENTIC TRADING
Quant's financial agents will run on AOS, Unicity's agent operating system - making security and compliance a property of the runtime, not a promise of the model
ARCHITECTURE EATS POLICY FOR BREAKFAST
Prompt injection spent two years as the risk at the back of most AI security discussions. In 2026 it became operational, with three separate disclosures against Claude that shared one property: the enforcement lived inside the thing being attacked.