nextup / engine

Ingest. Model.
Master.

Under the hood, NextUp builds a live model of your mind and continuously optimizes the shortest path from where you are to where you need to be.

01

Ingest

Syllabi, PDFs, slides, lectures, handwritten notes, problem sets. Everything parsed, embedded, and linked in seconds.

  • Multimodal parsing
  • OCR + LaTeX
  • Audio transcription
02

Model

A personal cognitive model tracks mastery, decay, and misconceptions for every concept, updated after each interaction.

  • Bayesian knowledge tracing
  • Forgetting curves
  • Misconception detection
03

Master

Agents generate the exact drill, explanation, or challenge you need next. Not more studying. Better studying.

  • Adaptive problem gen
  • Optimal scheduling
  • Exam simulation
// Spec sheet

The numbers
we're building to.

Design targets for the NextUp engine. We'd rather aim absurdly high and publish the bar.

0sTutor first-token latency
0k+Concepts per personal graph
0sLecture-to-notes turnaround
0/7Background agents on your queue
0+File types ingested
0:1Model per student, never shared
// Principles

Your mind.
Your data.

◆

Private by design

Your cognitive model belongs to you. Built to be exportable, deletable, and never sold.

✦

Honest about uncertainty

Every forecast shows its confidence. The tutor tells you when it isn't sure.

◉

Learning over shortcuts

The engine optimizes for what you can do without it, not for how much you lean on it.

"The future student won't study harder than everyone else. They'll have a better operating system."

— the NextUp manifesto
Next layerRoadmap →