Assess work that belongs to the role.
Define the engineering competencies before choosing the task. State what the assessment covers, and what it cannot establish.
Trust & methodology
Better evidence starts with honest boundaries.
A compelling report is not the same as a validated assessment. Our evaluation principles put inspectable work, explicit limits, and human accountability first.
A finding needs supporting evidence.
Connect interpretations to code, executed checks, or follow-up excerpts. A polished explanation from an evaluator is not evidence by itself.
“Not established” is a useful answer.
Distinguish a demonstrated gap from missing evidence, an interrupted session, or a task that never exercised the competency.
Support decisions. Don’t automate rejection.
Hiring teams retain responsibility for interpretation and final decisions. Uncertain or adverse findings need an appropriate human-review path.
Consistency is a goal to test.
Job alignment and structured criteria matter, but do not establish that a system is bias-free. Evaluate agreement, repeatability, and outcome patterns.
Make the experience transparent.
Explain permitted AI use, collected evidence, data handling, and support options before the candidate starts. Avoid hidden standards.
What this does not claim
Engineering evidence. Not mind reading.
The evaluation does not infer personality, emotion, protected characteristics, or “culture fit”, and it never analyses a face or a voice. AI use alone is not evidence of misconduct. A screen is not a guarantee of future job performance.
Where each policy stands
Read the policies. Then invite candidates.
We make no certification or compliance claim we cannot back. Here is where each one stands today.
The security pageCandidate data and recording
Before they start, candidates acknowledge a versioned notice of exactly what is kept: their written plan, code and its history, test runs, terminal commands, AI exchanges and the interview transcript. The interview is recorded — screen, camera and microphone — and the coding is not filmed. No recording reaches scoring, which reads the transcript alone. A written retention schedule is not yet published.
Model providers and data use
The model, voice, sandbox and hosting providers are listed by name on the security page. We claim nothing about how a provider handles data beyond what its terms and our settings support.
Security and enterprise review
We do not hold a SOC 2 report or an ISO 27001 certification. The security page lists the controls that do exist, with the values they run on, and the documents we cannot hand you yet.
Accessibility and accommodations
There is no published accommodation process yet. A candidate who needs one can write to [email protected] before starting. The accessibility page says where the site and the assessment stand.
Bring your questions. We’ll start with the evidence.
Use a walkthrough to review product fit, methodology, and pilot requirements.