Calculated Labs
Tutor V3 · For curriculum providers & schools
Verrow Tutor V3 gives curriculum providers a working foundation for guided AI tutoring connected to their materials. Students can receive explanations and help with their own work, while parents and educators can review the support provided.
Available for partner evaluation. V3 includes live, curriculum-grounded tutoring and a reviewable Learning Ledger. Learning Layer and Curriculum Intelligence demonstrations use reviewed synthetic examples. The current Mac build is an evaluation candidate.
Current V3 evaluation screens. Tutoring examples use evaluator-written prompts and live responses. Select any image to look closer.
Your curriculum provides the instruction. Verrow extends that instruction with guided tutoring when students need another explanation, a helpful question, or feedback on their own work.
Verrow is designed to work with your selected curriculum, respect educational boundaries, preserve student responsibility, and provide adults with visibility into the assistance given.
For your organization, this creates several opportunities.
Families could receive guided tutoring connected to the materials they already use. This additional support gives prospective customers another reason to consider your program and creates opportunities for enhanced curriculum packages, memberships, or optional paid services.
A family struggling with a lesson does not necessarily need a different curriculum. Sometimes a student simply needs another explanation, a helpful question, or guidance toward the next step.
Verrow provides a way to offer that assistance within your educational approach.
Helping families work through these difficulties could improve their experience, strengthen confidence in your program, and give them another reason to continue.
Verrow can explain concepts, guide approaches, and respond to students’ own work using selected curriculum materials.
A partner deployment could make this assistance available beyond normal support hours, subject to service availability.
This creates an opportunity to support more learners without proportionally expanding individual academic support staffing, while reserving your team’s time for situations requiring personal attention, specialized expertise, or human judgment.
A practical starting point
Complimentary
Request the download, invite code, and setup instructions. Try your own lesson or the included Constitution sample at your own pace.
Request an evaluationAvailable for Apple Silicon Macs. Internet access is required for live tutoring. The Windows build is complete and awaiting final testing. Email contact@calculatedlabs.com for access.
$3,500–$7,500
A Compatibility Sprint can then explore your educational requirements, preferred delivery model, and practical path toward implementation.
Preliminary range. Pricing depends on curriculum, users, deployment, integrations, privacy requirements, and support.
Pilots, production deployments, and white-label options are scoped to each partner.
Discuss a Compatibility SprintA Self-Guided Experience
No script or technical experience is required. After receiving the download and invite code, the built-in walkthrough leads an evaluator from curriculum selection to Parent guidance, live tutoring, the Learning Ledger, the Learning Layer, and Curriculum Intelligence.
Use the short route first. Open the full guide only when you want more detail.
Explore at your own pace
Open a topic for the details and examples.
Curriculum Authority
Verrow is the guide.
Providers choose the material and its source boundary. In Curriculum Grounded, Verrow uses the selected curriculum as the academic authority for tutoring. It provides the model with relevant passages from those materials and instructs it to stay within them.
The qualified tutoring configuration uses GPT-4.1 in Curriculum Grounded mode. Other source modes are visible in the evaluation, but are not covered by the same qualification. Source boundaries guide the model; they do not guarantee an error-free answer.
Curriculum selection is explicit.The provider can see which materials are active, switch to their own material without mixing sources, and choose how closely Verrow must stay within them.
Real Tutoring and Student Ownership
Verrow can explain a concept, define a term, compare ideas, model a bounded step, help plan an approach, and give feedback on a learner’s own work. It uses the selected curriculum and the conversation to offer useful help while preserving student authorship.
The learner asks how Congress can check a presidential veto. Verrow explains the process from the Constitution material, then checks the learner’s own explanation and recognizes what they have understood.
Guidance that responds to the learner.Live V3 responses to evaluator-written prompts. Verrow explains the relevant source, then acknowledges the learner’s reasoning without writing a finished submission.
If a learner asks for a finished answer to hand in, Verrow maintains the authorship boundary. When the learner asks for help organizing their own explanation, tutoring can continue.
Ownership is protected without ending the learning.Verrow declines to write the submission, then helps the learner organize an explanation in their own words.
Parent and Educator Guidance
Parent View brings age guidance, pacing preferences, homework boundaries, writing support, and family guidance into one place. Current tutoring qualification covers age-aware support and core student-authorship protections; individual preferences should also be tested with a partner’s lessons.
The evaluation lets you move freely between views. Protected adult access and role permissions must be configured and validated for a production deployment.
Clear choices in plain language.The Parent View exposes guidance choices for adults to review and evaluate alongside the tutoring.
The Learning Ledger
The Learning Ledger records the delivered response alongside source use, tutoring mode, and the recorded decisions behind the help. Adults can inspect the conversation, a parent summary, and more detailed evidence of how AI contributed.
This Ledger records the learner’s reasoning and Verrow’s feedback from the exchange shown above. The parent summary identifies the curriculum source and explains how AI contributed.
Parents and educators can review the support while the learner continues thinking. Automatic summaries and labels can be mistaken; the transcript remains available to check what was actually said.
Immediate value for parents and teachers.The same tutoring turn is visible alongside its source and parent summary, so adults can see the help the learner actually received.
The Learning Layer
An answer reached with hints is different from an independent explanation. V3 can record compact learning observations locally after eligible tutoring turns. Its Learning Layer demonstration uses two reviewed synthetic histories to show different proposed starting points, keeping the earlier support visible.
In this synthetic history, the learner reached the idea with choices and a reminder. The proposed next step reconnects the branch roles before asking the learner to explain.
In this synthetic history, the learner later explained the idea independently. The proposed next step offers less scaffolding and checks whether the understanding transfers.
Verrow observes evidence of learning. It does not claim certainty about learning.
How this is demonstrated: The screenshots show proposed next steps from synthetic evidence. Reliable ongoing adaptation, long-term progression, and mastery have not been established by this demonstration.
Curriculum Intelligence
Understanding where learners need additional support could help your team decide where to add an explanation, example, or supplemental resource. V3 currently demonstrates this direction through reviewed synthetic examples; live provider analytics remain future work.
The current synthetic example shows how many curriculum mappings have been reviewed, with the count and its denominator kept together. It distinguishes available evidence, privacy-suppressed results, and unavailable information, then separates what was observed from what may merit review and what has not been established.
Measures of recurring learning difficulty and curriculum-version comparisons are not established capabilities of this demonstration.
A visible path from evidence to review.The synthetic display illustrates eligibility rules and aggregate evidence. It contains no real learner records or raw tutoring content.
Useful without overclaiming.The provider sees the available count in context, the question it may raise, and the conclusions the evidence cannot support. Privacy-suppressed and unavailable results never become misleading zeroes.
Privacy, Control, and Readiness
The current evaluation uses protected hosted GPT-4.1 access. Production decisions about accounts, course access, curriculum delivery, integrations, hosting, retention, and validation are scoped with each partner.
Uploaded materials, session records, chat history, and Learning Ledger history are stored on the evaluation computer. For a live answer, the request and limited lesson or guidance context travel through Verrow’s protected service to the AI provider. Provider credentials stay on the server. Local storage does not mean offline tutoring or that no content leaves the device.
The accepted tutoring release passed 3,596 automated checks and a 100-turn live-model evaluation across five new curriculum packs. Responses and recorded source evidence were reviewed by Codex. The evaluation included continuous assignments, adversarial challenges, and different subjects and source conditions.
Recognized urgent-safety messages interrupt tutoring and direct the learner toward adult or appropriate emergency help. This safeguard does not recognize every possible risk or replace adult oversight.
This is internal testing with evaluator-written scenarios and AI-assisted review, not independent human validation, third-party certification, or evidence of student outcomes. The 100-turn evaluation retained two ordinary tutoring errors. AI responses and automatic summaries still require judgment, and each partner’s curriculum and learner population need their own validation.
Page reviewed against V3 3.0.0-demo.20 on 5 October 2026. The tutoring evidence above comes from the accepted release underlying this presentation update.