Learn it.
Prove it.
Semideus turns what you study into durable understanding through teach-back, retrieval practice, and spaced review. Then it grades you. Your Mastery Score is evidence, not a streak.
Every learning app measures activity. Semideus measures understanding.
The illusion of knowing.
Streaks, completions, and highlights measure activity, not understanding. Rereading feels like learning. It is not. And AI makes the illusion worse by doing the thinking for you.
feels like knowing
is knowing
Built on how memory actually works.
Capture anything into knowledge cards and a wiki knowledge graph. Study through teach-back, retrieval practice, and spaced repetition the system schedules for you.
- Drop PDFs, papers, links, or repos. Cards and wiki nodes come from your sources.
- Every claim traces back to a source passage.
- The schedule adapts to what you get wrong.
Teach it back. Get graded.
Explain concepts in your own words. Semideus grades your explanation server side across four dimensions. No self-report, no participation trophies.
Attention masking hides future tokens so the model cannot peek ahead. During training, each position only sees what came before it, like covering the next page of your notes while you recite this one.
You did not say why masking still matters at inference time.
Explain it in your own words. Claude grades it against your source and names the exact gaps.
Questions generated from your own sources, scored against a rubric.
A card counts when it survives spaced retrieval, not when you saw it once.
There is no mark-as-learned button.
A score you can show.
Your Mastery Score is backed by logged evidence, not streaks. Every point traces to a graded event you can open. Watch it honestly.
- Computed from proof events, never self-report
- Gaps surfaced in plain language
- Moves only when a grade lands
The test is the training.
Retrieval and calibration are the two things that reliably make people better learners. Every graded teach-back is both. The loop that proves your understanding is the same loop that improves it. That is learning how to learn, measured instead of promised.
Calibration
Predict how well you taught it before the grade lands. Semideus scores the gap.
Spaced retrieval
Cards come back on a schedule tuned to your forgetting. A long interval is the real test that memory held.
Nothing your AI teaches you evaporates.
Connect Claude or any MCP client. Say deposit what I just learned. Semideus turns the session into knowledge cards and wiki nodes automatically.
Then the loop takes over: the cards get scheduled, the concepts get taught back, and what your AI explained becomes something you can prove.
Scheduled for review. First retrieval due tomorrow.
Learning is free. Proof is what you pay for.
Not a bigger library. A higher Mastery Score you can stand behind.
Every plan includes monthly AI credits for teach-back, tests, and Socratic sessions. Top up with credit packs anytime. Pack credits never expire.