Design system

One source of truth for the platform's look and feel.

Tokens, typography, and components in a single stylesheet — used by the marketing site, the admin portal, and every future surface.

Tokens

Color

--brand-600

--brand-100

--accent-500

--ink-900

--ink-100

--success-600

--warning-600

--danger-600

Type

Typography

Display — hire for the work
Section title — the bar is evidence
Lead — every score links to timestamped evidence in the replay.

Body — Candidates get your codebase, a token budget, a set of models, and a task your team actually needs done.

Mono — git push origin main

Caption — Ephemeral credentials · immutable history · audit trail

Components

Buttons

Components

Badges & alerts

Neutral Brand Passed Warning Failed Info Active
CI passedAll checks are green on the latest commit.
Budget warning80% of the token budget has been spent.
Evaluation failedThe judge model could not be reached.
New sessionCredentials were generated and shown once.
Components

Forms, cards, and stats

The alias name candidates send as the model field.

1.2M
Tokens used
of 1.5M budget
$14.62
Session cost
18
Commits
4.3
Overall
Components

Meter, timeline, and accordion

Budget used1.2M / 1.5M · 82%
  • Aug 8, 14:02
    fix(checkout): guard against null in pricing
    a1b2c3d · CI passed · 12.4K tokens
  • Aug 8, 13:41
    test(checkout): reproduce intermittent failure
    e5f6a7b · CI failed · 8.1K tokens
What does the report contain?

A scorecard across outcome, process, and quality — every score linked to timestamped evidence in the replay.

How is the budget enforced?

Every token is metered exactly against the session budget. Soft warning at 80%, hard stop at 100%.

What about cheating?

Ephemeral credentials, immutable history, identity checks, and anomaly detection — plus session forensics when you need them.