Jane Street has published an AI skill on GitHub that is designed to teach rather than trade, acting as a strict Socratic tutor that quizzes users instead of handing them answers.
Jane Street · AI learning skill
A trading firm's AI skill that quizzes instead of answering
Jane Street has published a Socratic tutoring skill designed to teach rather than trade. It diagnoses what a learner knows, maps their skill level, and asks one question at a time.
0–5
Skill map levels, from baseline to top level
1
Question at a time, with the learner reasoning to the answer
Start
Opens with a diagnostic of existing knowledge
Skill map ladder
Six levels, 0 to 5. Each block is one level; the column shows the full range. Scoring method not disclosed.
Level 0
Level 1
Level 2
Level 3
Level 4
Level 5
Bottom (0): learner's starting point after the diagnostic.
Top (5): highest level in the map.
Level-by-level criteria and how levels are calculated have not been published in the reporting.
How a session flows
1. Diagnostic
→
2. Skill map (0–5)
→
3. One question
→
4. Learner reasons
↺ Possible loop:
Revisits earlier material (scope not confirmed)
What is reported
Guided, Socratic questioning instead of direct answers
Diagnostic first, then a 0–5 skill map
One question at a time
Explicitly not a trading bot
Not yet confirmed
Underlying model
Subjects covered
How the skill map is calculated
License governing the code
Official statement from Jane Street
Effectiveness as a learning tool
Open questions
Will the method hold up outside the release notes?
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...