What the studio is thinking about
Short pieces from the engineers, designers and operators on the team. No product announcements, no SEO bait.
Field notesWhy we write the eval suite before we write the agent
On the discipline of starting from a scoring function — and the awkward first week where the prototype scores 31%.
EngineeringRouting between frontier and open models without losing sleep
A small piece of plumbing that decides which model gets which call. Saves money, ages well, doesn't get cute.
Case studyWhat 3.8M conversations taught us about ticket triage
The tickets that look easy are the ones that bite. Notes from a year of production triage.
Field notesThe agents we retired in 2025 (and what we learned)
About one in five prototypes never makes it to prod. Here are the patterns in the ones that didn't.
OperatingHow we onboard with security teams without losing weeks
The security review is the project. Treat it like one and the rollout gets a lot quieter.
EngineeringTelemetry for agents: what we log and why
A look at the shape of our trace schema, the dashboards we keep, and the alerts we don't bother with.
Have a workflow that deserves an agent?
Tell us what's eating your team's afternoons. We'll come back inside three days with a discovery plan, a price, and the names of the engineers we'd put on it.
