Evidence of how the paper was written.

One thing stops: creating a new class. That's it. Everything you already built keeps working, permanently — your existing classes still accept submissions, you can still create assignments inside them, grading and CSV export are unchanged, and every certificate ever issued stays verifiable forever (verification runs in the browser and never checked an account in the first place). Nothing is deleted, locked, hidden, or downgraded, and no student loses access to their own work. The free window runs 120 days from signup, which is longer than a 16-week semester plus exams — so a first course finishes inside it no matter when in the term you start. When you want to run another term, email us.

Because a detector and Solus answer different questions. A detector reads the finished text and estimates how likely it is that a machine produced it — a probability, generated after the fact, with documented false positives on English-language learners and formal academic prose. Solus records the writing while it happens and produces a signed, re-checkable record of it. When a student disputes a detector score, you have a percentage and a professional judgment; when a student disputes a Solus certificate, you have the ledger, and so does the student. Plenty of schools keep both — the detector as a triage signal, Solus as the record that survives an appeal. See the full comparison →

Network blocking and lockdown browsers work on your network, on your devices, during your hours. Take-home writing happens on a personal laptop over a home connection or a phone hotspot, and that's where most of the essays — and most of the AI — actually are. A blocker also produces no positive evidence: a student who never triggered a blocked request looks identical to one who used AI somewhere you can't see. Solus travels with the document instead of the network, so it works the same whether the student wrote in your classroom or at their kitchen table.

The cryptographic half (20 checks) works on the very first document a student writes — no setup, no history needed. The behavioral half compares this submission against that student's own typing baseline, and a baseline has to be built from their real writing first. Solus needs 3 qualifying sessions, across 3 separate days, totaling about 8,000 accepted keystrokes — for most students that's the first two or three assignments. Until then the behavioral checks report Enrolling, shown in amber, and the verdict reads "provenance verified, writer enrollment in progress." That amber is not a warning about the student; it means the comparison isn't available yet. Both the student's editor and your dashboard show a live progress meter for exactly how much is left, so you always know when the badge will unlock.

No — and we couldn't if we wanted to. Essays are encrypted in the student's browser before anything is transmitted, with a key wrapped for you; our servers only ever hold ciphertext. We don't train models, we don't sell or broker data, we don't run ads or third-party trackers, and none of this is a setting you have to find and switch off. Institutions can sign a FERPA Data Processing Agreement and export or delete their data on request. Privacy policy →

This is the honest hard case, and it's the one the paste blocker genuinely cannot catch — the keystrokes are real. It's handled on the behavioral axis instead. Transcription and composition leave different traces: where pauses fall relative to word and sentence boundaries, how often the writer goes back and revises mid-document, how bursty the typing is, how attention moves in and out of the window. Solus scores those four signals into an anti-transcription composite and recomputes it from the raw ledger at verification time, so a modified certificate can't fake the number. How the anti-transcription check works →

The certificate is cryptographically signed — it can't be selectively altered. If the signature verifies, the data is authentic. If the student claims their low keystroke count or high paste count is a bug, you can ask them to demonstrate the bug occurring in a new document. Solus logs are deterministic: a document written by a human will always produce a plausible keystroke pattern. A low-keystroke document with high paste volume is what it looks like.

Yes. The public verifier at verify.soluseditor.com (or the Verify page on this site) lets anyone drag and drop a .solus file to run the full 30-check verification — no account required. Everything runs entirely in your browser using the SubtleCrypto API. Nothing is uploaded.

Five possible outcomes. Verified Human™ — both axes green; the badge is awarded. Partial — provenance is intact but one or more behavioral checks came back mixed (borderline composition score, low keystroke count for the word count, etc.) — review in context. Not yet available — file is cryptographically clean but the writer has no baseline yet, so the behavioral axis can't run — common for legacy files and new writers, and an absence of evidence rather than evidence against. Baseline building — provenance intact, the writer is partway to an established baseline; the badge unlocks once enrollment completes. Tampered — a cryptographic check failed; the file was modified after signing. Treat this last one as a serious finding.

No. Solus works offline for writing and local saving. The keystroke signing is entirely local — it uses the device's cryptographic hardware, not a server. Students only need internet to submit to your class dashboard. They can write offline and submit later.

Yes. Students can open the local verifier in Solus to see their own certificate before submitting. This is intentional — we believe transparency builds trust. Students know exactly what you'll see. There are no surprises.

Document content is encrypted with AES-256-GCM. X Point's servers never see the plaintext of essays. You (the teacher) can decrypt and read the essay through your dashboard because you hold the class decryption key. But X Point cannot read your students' essays — only you can.