Skip to main content
A Human Signal Sprint is a structured evaluation run. You describe your model, its intended use case, and the population it’s supposed to serve — CoLinear handles everything from there. Scenario generation, contributor assignment, inter-annotator agreement, and Score synthesis all happen without you managing a single task. What you put into intake is what shapes the quality of your Score, so it’s worth doing carefully.

​
The Entry Flow

1

Choose Your Membership Tier and Complete Checkout

Select the tier — Validate, Refine, or Scale — that matches your model’s maturity and your evaluation goals. Complete Stripe Checkout to activate your membership. No sales call is required for Validate or Refine.
2

Receive Your Magic Link

After payment, you’ll receive a magic link via email from Resend. This link is your login — there are no passwords in CoLinear. Click the link to access your dashboard and begin your sprint.
3

Complete the Waiver

Before submitting intake, you’ll complete a short waiver covering data handling, contributor confidentiality, and acceptable use. This takes less than two minutes.
4

Complete the Sprint Intake

The intake form is where you tell CoLinear what it needs to generate strong evaluation scenarios. You’ll describe your model, its intended use case, your target user population, and what you specifically need to learn from this evaluation.
5

Submit Your Sprint

Submit the form to enter the review queue. Your sprint status moves to intake_submitted and automated Gate 1 begins immediately.

​
What Happens After You Submit

Your sprint passes through two quality gates before contributors see it. You don’t need to manage either — but knowing what they check will help you write intake that clears them cleanly.
An automated quality check runs immediately after submission. It checks for:
  • Completeness — all required intake fields are present and substantive
  • Valid sprint type — your evaluation type is supported on your current membership tier
  • Valid contributor count — your requested contributor count must be one of the supported values: 10, 15, or 25
  • Valid turnaround window — your requested turnaround must be one of the supported options: 5, 7, or 10 days
  • Sensitivity flags — your sprint is checked against categories that require mandatory admin review: health, finance, legal, and minors
If Gate 1 passes, your sprint moves to task_drafted and an AI-generated contributor task brief is created for admin review.If Gate 1 finds issues, your sprint moves to needs_revision. You’ll see specific notes in your dashboard explaining what to fix. Edit and resubmit to re-enter the queue.
A member of the CosentriQ team reviews the AI-generated task brief before any contributor sees it. They may edit the brief for clarity, contributor-appropriateness, or scope. They may also request changes from you if the intake needs more specificity.You do not write contributor-facing tasks. CoLinear translates your intake into the task structure that produces clean, consistent contributor evaluations. This is intentional — contributor task quality is one of the primary drivers of Score reliability.Once the admin approves, your sprint moves to approved and is queued for contributor launch.

​
What to Include in Your Intake

The specificity of your intake directly determines the relevance of the scenarios CoLinear generates. Generic input produces generic scenarios — and generic scenarios produce wide confidence bands that are harder to act on. Write strong intake by including:
  • A clear hypothesis about your model’s job: Not “it answers questions” — more like “it triages inbound customer support requests and routes them to the correct resolution path without human intervention.”
  • A specific target user population: Not “enterprise users” — describe the role, the context, the prior AI experience, and the stakes involved. A financial analyst who uses three data tools daily reads model output differently than a branch manager who doesn’t.
  • Your model’s claimed behavior: Be precise about what your model claims to do well. If you’re testing a summarization model, specify what it claims to preserve — tone, key decisions, action items, or all three.
  • Known weaknesses or edge cases you want tested: If you already suspect your model struggles with negations, multi-step reasoning, or ambiguous requests, say so. CoLinear will generate scenarios that probe those areas.
Better intake = tighter Score. The more specific you are about your target population and model claims, the more relevant the generated scenarios will be — and the narrower your confidence bands at the same sample size.

​
Sensitivity Flags

If your sprint involves any of the following categories, it will be automatically flagged for mandatory admin review at Gate 2 regardless of your membership tier:
  • Health — clinical decision support, patient-facing outputs, diagnostic or triage use cases
  • Finance — investment recommendations, credit decisions, financial planning outputs
  • Legal — contract review, legal advice, compliance outputs
  • Minors — any model with a target population that includes users under 18
A sensitivity flag is not a rejection. Your sprint continues through the process — it simply requires explicit admin review and sign-off before going live. Flagged sprints may take 1–2 additional business days at Gate 2.