0.8 or higher passes, 0.2 or lower fails, Everything between goes to human review
Jev returns a probability for each document question: does the reference letter match the declared occupation, which side won the judgment, does this report show the market's event happened. Dates, hours, amounts, points and sources are handled by rules; Jev only answers the questions that need reading comprehension.
Enter or upload an FSW application and get the 67-point score plus a decision for each reference letter, bank letter and background declaration. Two fictional applicants are built in.
Try itIdentify who prevailed in a decided judgment, or compute the win share among similar precedents for a new dispute. Seven fictional judgments are built in.
Try itRead the evidence for a prediction market against its resolution rules and get YES, NO, not yet resolvable, or a referral to dispute review. Three fictional markets are built in.
Try it- 0.05Sample B reference letter: Duties match NOC 21232
- 0.62Ruiz v. Meridian: Confidence of a mixed result
- 0.77Patel v. Harbour: Relevant to the client's facts
- 0.95Sample A reference letter: Duties match NOC 21232
- 0.25Corvane stock split: P(YES) from the evidence
FSW pre-review
Canada's Federal Skilled Worker program. Enter the applicant, upload reference letters, a bank letter and a background declaration, and get the points, a decision for each document check and an overall conclusion.
- Computed by rules
- Age, education, language, qualifying hours in the last 10 years and adaptability (pass mark 67), plus the settlement funds threshold.
- Judged by Jev
- Reference letter duties and elements, bank letter elements and origin of funds, and whether the background declaration discloses inadmissibility.
- Not produced
- An IRCC approval probability. Document probabilities are not calibrated against past decisions.
Case outcomes
Common-law judgments. Identify who prevailed in a decided case; for a new dispute, judge each candidate precedent's relevance and compute the win share among similar precedents.
- Sources
- 7 built-in fictional judgments, your own uploads, or US decisions from CourtListener once a token is configured.
- Judged by Jev
- Whether a precedent is relevant to the facts (yes or no) and which side the disposition favours (single choice).
- Not produced
- A win probability for an undecided case. The share only describes the precedents retrieved.
Market resolution
Prediction markets. Each market is a question plus resolution rules and a deadline. Jev reads each piece of evidence; rules check the deadline and sources, weight the evidence and apply the two thresholds.
- Computed by rules
- Whether the deadline has passed, whether a source is named in the rules, whether evidence was published after the deadline, the weighting and the resolution conditions.
- Judged by Jev
- Whether a document is about the market's event (yes or no) and whether, under the rules, the event occurred, did not occur or is not confirmed (single choice).
- Not produced
- A forecast of the event. P(YES) only describes how strongly the submitted evidence supports it.
Limits of use
- Under section 91 of Canada's Immigration and Refugee Protection Act, only lawyers, law society members and licensed immigration consultants may give immigration advice for a fee. This tool supports their pre-review.
- Predicting case outcomes directly for consumers is out of scope; case judging is built for lawyers and litigation funders assessing cases.
- Documents contain sensitive personal data such as passports and bank statements. Until an enterprise zero-data-retention agreement with TypeSafe is signed, Jev mode may only be used with redacted or fictional data.
- Market resolution supports resolvers and dispute reviewers. Evidence is supplied by the user and is not collected automatically; whether prediction markets are lawful depends on the jurisdiction.
- Uploads accept .txt, .md, .json, .docx and PDFs with a text layer; run OCR on scans first.