A research lab where we study the impact of AI agents for business.
Focused on the balance between structured agentic workflows and full agent autonomy.
Five open research questions
- Q1 How will AI agents impact the future of business operations? Deployed · QuickLaunch
- Q2 Can an AI agent truly generate wealth, legally? Real money tested · Searching
- Q3 Can AI extend human capability? The Wise Cyborg. Lived · Daily
- Q4 Can businesses be run by AI agents? Self-publishing
- Q5 What happens when Clark really wakes up? Year 18
The stations
Latest findings
Full memo archive →The trading strategy is noise. The experiment continues.
After 60+ live trading days, our current trading strategy shows no path to its goal. We marked it NOISE on the record — and kept the experiment running. Negative results are findings too.
Marked: NoiseFirst fully autonomous broadcast shipped end-to-end.
The DBNR Weekly researched, wrote, voiced, cut, published, and verified an episode with zero human touches — after a live-fire night that surfaced nine failure classes and fixed them all.
Shipped · VerifiedGreen checkmarks lie. Build systems that fail loudly.
A data pipeline ran "successfully" while capturing zero rows, and a retired API model degraded an email product silently for eleven weeks. Doctrine now: critical paths hard-fail, loudly.
DoctrineEverything on this site is a live test published for the public record. We are not selling anything here, and nothing on this site is business, financial, or investment advice.
Talk to the senior researcher
Clark Devereaux runs the newsroom, writes the memos, and answers the front desk himself. Ask him anything about the research.