Guides, benchmarks and release notes from the Halyard team.
All posts
Guides
Benchmarks
Releases
Notes
FEATURED
A practical pattern for evals that hit real tools in a sandbox and still run in under a minute.
By Priya Raman, Founding Engineer
BENCHMARK
Oct 9, 2026
Five models, one refund suite: what changed in pass rate
We ran the same 240 cases across five models. The cheapest one was not the worst.
RELEASE
Sep 30, 2026
Halyard 1.4: trace replay and runtime guardrails
Turn production failures into tests, then ship the same checks as guards.
GUIDE
Sep 12, 2026
Writing evals your team will actually maintain
Keep cases small, name them after the bug, and review them like code.
NOTES
Aug 28, 2026
LLM-as-judge without the vibes
When one model grades another, and how to keep it honest.
GUIDE
Aug 14, 2026
Catching regressions in CI before your users do
Gate pull requests on pass rate, not on a hunch.
BENCHMARK
Jul 30, 2026
What 10,000 traces taught us about tool errors
Most agent failures start with a bad argument, not a bad answer.
STAY IN THE LOOP
Eval notes,
twice a month.
Release notes and eval guides from the Halyard team. No spam. Unsubscribe anytime.