HealthcareJuly 14, 2026 · 12 min
The CMS-0057-F engineering checklist: four APIs, six standards, one date
The sequencing that actually works for the January 2027 deadline, and the three places payer implementations usually stall.
Coming SoonInsights
Written by the engineers who do the work.
Field notes on building and proving enterprise AI — evaluation, integration, and the regulatory deadlines that shape our clients' roadmaps. No ghostwriters, no listicles.
EvaluationJuly 28, 2026 · 9 min
Nearly a third of teams run no evals. Here's what the other two thirds get wrong.
Human spot-checks feel like evaluation. They aren’t. A walk through the failure modes we see in eval suites teams built themselves.
Read Article →29.5%
of teams run no evaluations at all
0.71 → 0.87
what one rubric revision did to inter-annotator agreement
RegulationJune 30, 2026 · 8 min
EU AI Act dates that matter if you ship AI into Europe
August 2026, December 2026, December 2027, August 2028 — what each deadline covers, and why credit scoring teams have the least time.
Coming Soon
LogisticsJune 12, 2026 · 7 min
Two carrier portals had no API. What we did instead of browser automation
Scheduled file exports are unglamorous and dramatically less brittle. Field notes from the 3PL invoice-audit build.
Coming Soon
EvaluationMay 27, 2026 · 10 min
LLM-as-judge is fine. Uncalibrated LLM-as-judge is a rumor with a dashboard
How we calibrate judges against human review on a golden set, and why we report the agreement number instead of asserting it.
Coming Soon
RegulationApril 18, 2026 · 6 min
Your DORA register probably stops one level too early
Incomplete registers of information drew the most supervisory attention in the first enforcement cycle.
Coming SoonGet the next one by email.
One email when we publish. No digest, no drip campaign, unsubscribe works the first time.