Public overviews of our subject areas.
Four threads run through our work. Here we give a taste; the depth — articles, slides, and the discussions behind them — lives in the members area.
How is software produced like a factory — repeatable, auditable, and with governance built in rather than bolted on? The idea is not a smarter script but a FactoryOS: a project layer that specifies what gets built and a process layer that runs it as chains of specialised agents, each turning defined inputs into defined artifacts. We share reference architectures, deployment pipelines, and harness-agnostic platform blueprints drawn from real enterprise settings — and show where the concept transitions from experimental tech into infrastructure that holds up in production.
Agent-OS systems: orchestrating multiple agents, tool integration, memory, and guardrails. The harder question is not the demo but the operating model — how decentralised agentic systems scale inside human organisations, what holds up under production load, and what breaks. We work vendor-neutrally across the whole stack, from sovereign, self-hosted setups to retrieval-augmented systems in real use, judging each architecture through both a technological and an organisational lens.
Compliance, risk, auditability, and control for autonomous systems in the regulated enterprise — from policy-as-code to the question of accountability. Governance here also means quality: calibrated evaluation, LLM-as-judge methods, and peer review that turn “it seems to work” into a solid basis for continuous improvement. And it means the organisational dimension — how the automation of automation reshapes work — a conversation we deliberately hold across industry practice, research, and public administration.
Proven patterns and anti-patterns from real migrations — from monolith to factory. Concrete building blocks: agent chains and their artifact graphs, the split between project and process layers, and cases like autonomising the lean product process with a team of specialised agents. Through a rigorous peer-review loop we mature these from one-off tricks into repeatable methods — and stay honest about which shortcuts come back to bite.