DuetHire / Evidence flight recorder

AI-NATIVE ENGINEERING HIRING

Hire great engineers
who own what ships.

DuetHire turns an open engineering role into a calibrated work simulation. Candidates build with AI, adapt to changed requirements, debug flawed work, verify the result, and defend their decisions.

Synthetic session / TS-014

DuetHire

Would you ship their AI-written patch?

A 90-minute ownership simulation for senior engineering hiring—recorded from requirement change to human review.

ROLESenior backend
ENVIRONMENTSynthetic service
CASERelay / bounded retries
ELAPSED00:00
relay.tsSTARTER

            
DETERMINISTIC TEST OUTPUTNOT RUN

            

          


          
        

        
Illustrative synthetic data · AI use permitted and disclosed · No private prompts collected · Human decision required

      

    


    

      
ONE LONGITUDINAL STORY
A loop built around change.

      
    
        
  1. 01 / CALIBRATE
    Define the role and ownership signals.
    Set the competencies, permitted tools, synthetic task, deterministic checks, and reviewer rubric.
    2. 
        
  2. 02 / RUN
    Introduce consequential change.
    The candidate receives an at-least-once delivery constraint, surfaces a duplicate, and repairs the boundary.
    3. 
        
  3. 03 / DEFEND
    Probe the actual implementation.
    Adaptive questions test risk judgment and the limits of the candidate’s verified solution.
    4. 
        
  4. 04 / REVIEW
    Make evidence decision-ready.
    Artifacts, test states, explanations, notes, and open questions arrive in a human reviewer packet.
    5. 
      

    


    

      
REVIEWER PACKET / SYNTHETIC EXAMPLE
Judgment with a trail.
Every interpretation points back to observable work. No black-box score and no autonomous pass or fail.

      

        

        

      

    


    

      
EXPLICIT INPUTS. BOUNDED OUTPUTS.
Relevant without asking for secrets.

      

        
01
Approved context or synthetic abstraction
DuetHire never requires confidential repositories, tickets, incidents, complaints, code, or architecture.

        
02
Candidate-visible conditions
Permitted tools, time, recorded activity, evaluation intent, and data use are explicit before the session.

        
03
Evidence, not surveillance
DuetHire records task artifacts and deterministic outcomes. Private AI prompt histories are not collected.

        
04
Human decision boundary
Reviewer packets support peer review and hiring judgment. They do not make the decision.

      

      
ATS-NEUTRAL HANDOFF
DuetHire can sit alongside an existing applicant-tracking workflow for the information needed for invitations, status, and evidence handoff.

    


    
FAQ / BEFORE THE PILOT
Clear conditions from the start.
DuetHire is at manual pilot stage, focused on senior backend hiring.

      
What happens in the 90-minute simulation?
A feature task, requirement change, flawed AI-assisted patch or failing test, debugging and verification, then an adaptive technical defense—all inside a synthetic engineering service.

      
Can candidates use AI?
The loop defines and discloses permitted tools. DuetHire evaluates the technical work, explanations, verification, and judgment produced under those conditions—not private prompt history.

      
Do we have to share company-confidential material?
No. Use explicitly company-approved context or a synthetic abstraction. Private company material is not required.

      
What reaches the hiring team?
A reviewer packet with competency-aligned criteria, candidate artifacts, deterministic test results, explanations, reviewer notes, unresolved questions, and peer-review status.

      
Does DuetHire decide who passes?
No. DuetHire organizes evidence and decision support. Human reviewers retain the hiring decision.

    


    
MANUAL PILOT / SENIOR BACKEND
Design a loop that tests ownership.
Bring a role and company-approved context—or start entirely synthetic.
Design a calibrated loopOpens your email client · Example address must be replaced before publication