What It Is

Get Early Access

hwb2 Learning is an AI-augmented math tutoring platform I am actively building. It automates intake, first-pass feedback, and progress tracking so that human tutoring time is reserved for diagnosis, strategy, and explanation.

This is not “AI tutoring instead of a teacher.” The goal is to scale responsible, human-led instruction by removing repetitive overhead.


Problem → Insight → Solution

Problem

  • One-to-one tutoring does not scale
  • Administrative overhead consumes instructional time
  • Feedback can be delayed and inconsistent

Insight

A large portion of tutoring effort is repetitive cognitive labor that does not require human judgment: intake triage, rubric-like feedback patterns, and progress summaries.

AI is best used before and after sessions—not during—so the live session stays human and high-trust.

Solution

  • Standardized Google-based intake
  • AI-generated first-pass feedback (structured, reviewable)
  • Human-led tutoring focused on the highest leverage moments

Live Demo (v0)

Status: Limited scope, intentionally constrained.

Demo link: Coming soon (When ready, I will link a public v0 workflow here—likely an interest/intake form or a constrained “submit response → receive structured feedback” demo.)

What is real today

  • Repeatable ingestion workflow (intake → structured processing)
  • Feedback generation using a consistent template
  • Foundations for progress tracking and follow-up

What is still manual / in progress

  • Human review loop and refinement
  • Classroom-scale workflows and dashboards
  • Image/PDF ingestion

Architecture Snapshot

Current Stack

  • Google Forms → Google Sheets for intake
  • Python backend for processing and orchestration
  • OpenAI Responses API for structured feedback generation
  • Human review loop (manual at v0)

Planned Extensions

  • Image and PDF ingestion (including handwritten work)
  • Workflow automation (parent/student onboarding + follow-ups)
  • Student progress dashboards and reporting
  • Classroom-scale routines and data exports

Development Status

  • v0 ingestion pipeline: in active development
  • Building and iterating: weekly
  • Initial private alpha: planned
  • Public beta: TBD

Technical Details

Repository: Link coming soon (I will publish a clean repo once the v0 demo workflow is stable and appropriately scoped for public viewing.)

What the implementation emphasizes:

  • Secure API usage (keys, environments, minimal exposure)
  • Structured prompt design (repeatable, testable outputs)
  • Small-scope v0 iteration before scaling features

Build Log

  • Dec 2025 — Product framing + workflow planning
  • Jan 2026 — v0 ingestion + structured feedback iteration
  • Feb 2026 — Demo hardening + automation hooks (planned)
  • Spring 2026 — Image/PDF ingestion exploration (planned)