We ran TypeSafe's Jev, OpenAI's Decisions API and Cloudflare's Clef on 9,000 real Traba decisions, then on forged documents, tampered IDs, face matching and video. Jev wins on language at the lowest cost; none of them can do math.

Four core tenets of Traba engineering — and the four surest ways to fail each. What we screen for in an agentic world.

How Traba's evals grew up alongside its agents — from deploy-and-pray to operators and an agent fleet running the improvement loop themselves — to keep a fleet of production agents correct and safe as they make real decisions about real people.

Scout vets workers — but vetting is just one step in a much longer funnel. A new series on the agent infrastructure Traba is building to optimize the entire industrial worker lifecycle, and what “Physical AI” really means when your decisions affect real people's livelihoods.

Traba is reimagining industrial staffing by building Scout - an AI recruiter that conducts thousands of real-time interviews in parallel, qualifying workers with the speed and consistency of automation while preserving the depth and personalization of human-led conversations.

Traba transitioned from Firestore to PostgreSQL to manage escalating costs and improve scalability. The migration reduced database expenses by over 70% and enabled greater control and efficiency by switching to a fixed hourly billing model. Our journey highlights key insights for optimizing database architecture for growth.