Personalized practice
Identify a starting point, revisit prerequisites, and suggest the next activity using evidence of understanding.
ATHSOLVE / PRODUCT OVERVIEW
From a student's question
to a teacher's next step.
AI learning software in development for schools, students, and teaching teams.
INTENDED CUSTOMERSchools and teaching teams
END USERSStudents and educators
CURRENT STAGERunning development prototypes
THE PROBLEM
Answers don't always
reveal understanding.
A student can get an answer without understanding the method. A teacher can see a score without seeing where the student got stuck. Athsolve aims to connect guided learning with the evidence educators need to support it.
Guided tutoring
Use questions, hints, and explanations to help students work through a problem and explain their reasoning.
Teacher insight
Bring recurring difficulties and progress into view, with context for educator review and follow-up.
ILLUSTRATIVE PRODUCT WORKFLOW
One connected learning experience.
Set the learning context
A teaching team identifies a learning objective and selects relevant material.
Understand the student's attempt
The student works through a question and shares their approach.
Provide a useful next prompt
The intended tutor offers a targeted hint or explanation, followed by a check for understanding.
Support educator follow-up
A proposed teacher view summarizes the concepts practiced and difficulties observed. The educator decides what to do next.
This describes the intended product workflow. It is not a claim that all capabilities have been released.
CLAUDE-BASED TUTORING PROTOTYPE
Reasoning support,
with a clear learning purpose.
Our hint-ladder prototype uses the Claude API to turn a problem and a student’s attempted answer into four structured levels of support: a nudge, a concept reminder, a partial worked step, and a full solution. The aim is to reveal help progressively as the student needs it.
Inputs
Learning objectives, relevant course excerpts, and the student's question or attempt.
Model tasks
Generate a targeted hint, explain a concept, ask a follow-up question, or summarize a learning difficulty.
Evaluation
Check accuracy, alignment with source material, quality of hints, uncertainty handling, latency, and cost per session.
Development prototype: the Claude-based hint ladder runs separately from this public website. The website’s algebra example remains scripted and does not call the Claude API.
RUNNING DEVELOPMENT PROTOTYPES
A focused build.
Five connected components.
Our initial work brings guided mathematics practice and educator insight into a testable workflow. Contact the team to discuss a prototype demonstration.
Progressive hint ladder
Accepts a problem and student attempt, then uses Claude tool output to structure four levels of guidance. Support progresses from a nudge to a worked solution.
Explanations from lesson material
A prototype for explanations linked to teacher-supplied material, with section references and an intended refusal to go beyond the supplied evidence.
Session summaries
A structured view of concepts practiced, recurring errors, and suggested areas for teacher follow-up. Demonstration sessions are not evidence of a school pilot.
Misconception classification
Labels likely mistakes using a fixed taxonomy, including sign errors, inverse operations, fraction errors, and missed distribution.
A repeatable evaluation harness
Records early-hint answer matches, final-answer matches, misconception-label matches, latency, and token usage for a supplied test set.
Evaluation limits: the starter harness uses exact text matching. It can flag a number already present in a question, miss equivalent answers, or miss indirect answer disclosure. These checks support debugging; they do not establish pedagogical quality or mathematical correctness on their own. No benchmark scores or measured learning outcomes are claimed here.
CURRENT SCOPE & NEXT MILESTONES
What you can evaluate today.
Available on this website
- A product overview and proposed learning workflows.
- A scripted algebra walkthrough illustrating guided hints.
- A direct contact route to discuss prototype availability.
Next development priorities
- Expand and review the mathematics evaluation set.
- Improve checks for answer disclosure and equivalent answers.
- Refine grounded explanations and educator-facing summaries.
- Define an appropriate school pilot when ready.
A school pilot would require agreed learning goals, educator oversight, and data-handling arrangements. Pilot results and learning outcomes will only be published when they have been measured.
Bring us a learning challenge.
Discuss your school's goals and the prototype you would like to see.
Contact the Athsolve team