Local AI for lecture notes
A semester's slides, readings and notes become one searchable library that answers with citations. The studying stays yours; the losing track of things stops.
By week ten, a course is a pile: lecture slides, printed readings, typed notes, the occasional scanned handout, spread across folders and formats. A local AI turns that pile into a library you can question. Every document is extracted and searchable at once, answers arrive with citations back to the slide or page they came from, and nothing is uploaded anywhere. What the tool cannot do is the understanding. That remains the actual assignment.
The pile that a semester becomes.
Lecture material arrives in whatever format the week produced: slide decks as PDFs, journal readings, your own notes, a scanned handout from the one professor who still photocopies. A local tool reads the common formats on your machine, whatever the file type is, and scanned pages convert too, within the limits of how legible the handwriting is. Once imported, every week's material sits beside every other week's. The question "did we cover this in week four or week eight" becomes a search instead of a memory test.
Ask the semester, not the week.
Studying one lecture at a time hides the thing exams are built on, which is how the material connects. A grounded local tool answers across the whole library at once: where the definitions disagree, which readings contradict the slides, what the professor said that the textbook omits. The answer carries citations, so you read the claimed passage rather than trusting prose, and when the library holds no answer, the tool says so instead of filling the gap. That refusal is worth more than fluency during revision, because an invented fact in your notes is a wrong fact you now believe.
What stays yours.
The tool can retrieve and summarize; it cannot understand for you. Asking "explain the difference between these two theories" and reading a cited, correct paragraph is studying, but it is not yet knowing, and a student who outsources the reading outsources the learning. Use the tool to make sure nothing in the pile is lost and nothing in your notes is unsupported. The argument you build from it, the one the exam asks for, is the part only you can make.
How the loop maps to a semester.
Import each week's material as it arrives, so week eleven is never a catch-up problem. Ask the cross-lecture questions early, not the night before the exam, because an answer that names a gap is telling you what to go read. Keep your own notes beside the answers; in Istor they live in the same window as the sources they came from. And coursework stays local: nothing leaves the machine, which is the difference between a tool you can point at an unpublished thesis and one that needs a permissions conversation first.