What an accepted task costs
A month of agent build tasks priced by the task a reviewer accepted, and where a 3.6 times gap between two models came from.
I run a multi-agent workspace in production every day and publish what I measure on it: what the agents cost, how much of their work gets sent back, and which controls hold up. The architecture is open source and the essays are below.
Latest writing
A month of agent build tasks priced by the task a reviewer accepted, and where a 3.6 times gap between two models came from.
When AI does the work, the people checking it stop understanding it, and the research says a training course won’t fix that.
Here's a cheap way to test an AI change without a research team.
More writing
Projects
A redacted, world-readable snapshot of the workspace I depend on every day. It holds eighteen load-bearing patterns, an adoption guide, samples and an evaluation method. The whole architecture is mapped on one page.
I read other people's agent systems against the same eighteen patterns and publish what I find, including the corrections when I get one wrong. No pattern is present in all eight.
Tools
Instruments from the working system, all free. Everything runs in your browser, and your agents can call the same instruments over MCP at jamesross.ai/mcp.
Eighteen questions on how your agent workspace actually runs. Banded score across six dimensions, gaps ranked, fixes linked to the reference architecture.
Run the check →The organisation view. Twelve questions on whether your organisation is ready to put AI agents to real work, and which exposures to close first.
Assess readiness →Paste what you're about to publish. Catches emails, home paths, credential shapes and secret filenames before they ship. Your text never leaves the tab.
Check a draft →What a big early read really costs across a session: size times turns remaining, priced. Seeded with measured numbers from my own workspace.
Price a read →Two tracks through the same material. The modules explain why each capability matters, and a tutorial has you build a workspace yourself.
Start learning →About
I came to agents through data engineering, building production data platforms. That meant cloud data warehouses, ETL pipelines and the analytics executives decide on. Then I went deep on agent systems. I built a full multi-agent workspace from the ground up, and built and deployed an AI-driven SaaS end to end.
I bring an engineer's production discipline to it (version control, testing, provenance, rollback, observability), in a field that is still mostly demos. Separately, I'm a PhD candidate researching how autonomous agents share and govern memory.