Start with the submitted work
Use the essays, reports, files, code, repositories, and mixed assessments students already turn in. No course redesign is required.
WHAT YOU GET
A clearer way to review AI-era student work
Pruuva turns an existing assignment into a structured review: collect the work, ask the student to explain it, inspect the evidence, and keep a record that supports the final decision.
Keep AI suggestions inspectable
AI-assisted grading stays tied to rubrics, student responses, execution evidence, and confidence levels so you can inspect the basis for every recommendation.
Ask students to explain it
Submission-specific oral, video, or text follow-ups ask students to explain, adapt, and reason from the work they turned in.
Review the evidence in one place
See rubric-level understanding ratings, quoted transcript excerpts, integrity signals, strengths, weaknesses, and source-linked rationale in one report.
Make the decision with context
Use side-by-side submission review, evidence panels, rubric scoring, annotations, AI suggestions, and teacher overrides in one grading flow.
Preserve a defensible record
Each completed workflow can preserve identity, artifact, follow-up, rubric evidence, integrity context, and teacher judgment as a durable evidence event.
THE PROBLEM
Submitted work is no longer enough evidence
The AI-era question is not just who produced the output. It is whether the human can demonstrate the capability the output claims to represent.
AI made polished output cheap
Essays, reports, code, plans, and explanations can now be generated and revised faster than institutions can verify them.
Detection answers the wrong question
AI detectors guess whether text looks machine-generated. They do not show whether the person can explain the work, adapt it, or reason from it.
Capability evidence is expensive
A direct conversation, rubric review, session audit, and evidence record would be ideal for every student. Manually, that does not scale.
THE EVIDENCE LOOP
From submitted work to capability evidence
Pruuva is not an AI detector or a proctoring wrapper. It is a structured workflow for producing evidence that a human can defend the capability their work claims.
01
Artifact
What the student produced
The paper, project, file, code, repository, or written response becomes the starting evidence.
02
Probe
How they explain it
Adaptive follow-ups test reasoning, tradeoffs, corrections, transfer, and depth of understanding.
03
Rubric
How capability is evaluated
Findings map to rubric dimensions so evidence supports the grading decision instructors actually need to make.
04
Integrity
Whether the session is reliable
Consent, identity, timing, transcript quality, and session signals stay separate from comprehension findings.
05
Judgment
What the instructor decides
Pruuva suggests and organizes evidence. Teachers review, adjust, override, and finalize the grade.
06
Evidence event
What can be inspected later
A durable record links learner identity, task, artifact, probe, report, session context, and teacher judgment.
HOW IT WORKS
The loop from output to evidence
Pruuva brings the submitted work, adaptive follow-up, rubric evidence, and instructor review into one assessment flow. You get a clearer basis for grading without rebuilding the assignment.
- 01Define the task, rubric, and evidence policy
Create an assessment, choose supported artifact and question types, set rubric dimensions, configure access, and decide whether students demonstrate understanding through text, audio, or video.
- 02Students submit work and demonstrate capability
Students submit through a link or roster-gated portal. Pruuva probes specific claims, choices, code paths, sources, or methods so the student can explain the work in their own words.
- 03Review evidence and make the final judgment
The evidence report connects transcript excerpts, rubric ratings, integrity signals, and AI suggestions. Teachers review, annotate, override, grade, and export the record.
INSIDE THE ASSESSMENT
Not generic questions.
Evidence from their work.
Every probe starts from the student's own work: their claims, methods, sources, code paths, and decisions. Students respond through the mode the instructor configures, and the resulting evidence stays linked to rubric dimensions.
Reads the full artifactParses claims, evidence, methodology, reasoning, files, code decisions, or project structure.
Generates artifact-specific probesEach question targets concrete passages, choices, or claims with configurable depth and focus areas.
Captures demonstrated understandingStudents respond via the configured text, audio, or video mode with transcripts and session context.
Builds an evidence recordRubric ratings, quoted excerpts, integrity signals, strengths, weaknesses, and teacher decisions stay connected.
STUDENT SUBMISSION
The study employs a mixed-methods design integrating quantitative survey instruments with semi-structured interviews to triangulate findings.
Sample size determined via power analysis (α=.05, β=.80) yielding a minimum n of 84 per arm.
Findings consistently suggest that institutions with greater perceived credibility anchor expectations more effectively…
LIVE ASSESSMENT
PRUUVA
You mention a power analysis with α=.05, β=.80. Where did those parameters come from?
STUDENT
Um, alpha is just standard convention. The beta… honestly I think I pulled it from a template.
PRUUVA · FOLLOW-UP
Let's stay there. What does β=.80 actually mean in plain English?
STUDENT
It's the probability of detecting an effect if one is really there. Eighty percent.
PRUUVA · FOLLOW-UP
Good. So why .80 and not .90?
speaking…
A DIFFERENT APPROACH
Why this works when detection doesn't
FAQ
Things people ask
Most questions come from one place: making sure this is fair to students. We obsess about that too.
EARLY ACCESS
Stop guessing from output.
Start grading from evidence.
Build assessments that show what students can explain, defend, and apply — not just what they can submit.
Join early access