Suitorly.Lab
We're building models that measure what people can do. Not what they say about themselves.
This page is where we publish what we've built, what it scores, where it fails, and what we're doing about it. Updated when something changes, not on a marketing calendar.
Models
Five systems. One job each.
Every model has a model card, an eval set, a known-failure list and a human owner. Click one.
{{ m.k }}{{ m.ver }}
{{ m.name }}
{{ m.one }}
Model card · {{ mc.name }} {{ mc.ver }}
{{ mc.head }}
{{ mc.what }}
Inputs
{{ mc.inputs }}
Never sees
{{ mc.never }}
Built on
{{ mc.base }}
Owner
{{ mc.owner }}
Known failures
· {{ f }}
Evals · {{ mc.evalDate }}
{{ e.k }}{{ e.v }}
{{ e.s }}
Yellow mark is the release bar. A model that falls below it on any eval is rolled back, not patched in production.
Changelog · {{ mc.ver }} {{ mc.change }}
What we believe
Six things we'd rather say plainly.
These are commitments, not aspirations. If we break one, we'll write it up on this page.
{{ b.n }}
{{ b.t }}
{{ b.d }}
Lab notes
Short, signed, dated.
Alice Cheng
Founder. Writes these herself.
{{ n.date }} · {{ n.tag }}
{{ n.t }}
{{ n.lede }}
Read · {{ n.read }} →
What's next
The roadmap, including the parts we haven't solved.
{{ r.when }}
{{ r.t }}
{{ r.d }}
Open problem: {{ r.open }}
Research access
Check our work.
We give de-identified eval sets, scoring rubrics and fairness data to academics, regulators and clients' own data teams. Red-team the proof sessions. Tell us where they break.
{{ a.t }}{{ a.d }}{{ a.who }}
Request access
I want to
{{ w.t }}
Reviewed by a person within a week. lab@suitorly.com
Thanks, {{ sentName }}.
We'll read it properly and reply within a week. If you're red-teaming, we'll send the current proof-session variants and the disclosure terms first.