The thinking
behind the work.
Original perspectives on the data, decisions and human judgment that shape AI.
Read the latestThe foundations of useful AI
Data quality
A useful dataset is a maintained system.
The value of a dataset depends on what a team can explain, reproduce and improve after the first delivery.
Analysis · 6 min readModel evaluation
Evaluation starts with the decision.
A score becomes useful when it is tied to a specific workflow, a clear acceptance rule and the evidence behind each judgment.
Analysis · 6 min readHuman feedback
Human feedback needs a disagreement policy.
The useful signal is often in the reasons people disagree. A review process needs a way to preserve and resolve those differences.
Analysis · 6 min read