How LIORIN teaches — and how we validate it
LIORIN is a validated teaching engine built for instruction. This page explains the methods behind the engine, how it learns from review, the trust loop that keeps every move honest, and the validation program we use before making public performance claims.
WOPR-XII / LIORIN — IDENTITY
> WHO AM I .............. LIORIN, A TEACHING CHESS INTELLIGENCE > HOW I LEARNED ......... RECURSIVE SELF-PLAY + ENGINE-ASSISTED REVIEW > WHAT I DO ............. RUN ENGINE PRACTICE, REVIEW EACH MOVE > HOW YOU TRUST ME ...... PROPOSE / VALIDATE / EXPLAIN / LOG / REFLECT > LOCAL CHAT ........... DEMO ONLY, NOT SAVED OR MONITORED > BENCHMARK ............ FORMAL VALIDATION REQUIRED BEFORE PUBLIC CLAIMS > NOT A PROMISE ........ INSTRUCTION FIRST, OUTCOMES NEVER GUARANTEED
How LIORIN taught itself
LIORIN did not memorize a book of games. It learned by playing itself, millions of times over, in a recursive self-play loop. Each game is played, scored, and replayed — the positions where it lost or drew become the next lesson, and the cycle repeats.
Every self-play game is then run through engine-assisted evaluation: a second pass that grades each move, flags blunders, and confirms the winning continuation. The engine is the referee that keeps the learning honest, so LIORIN never rewards a bad habit that merely happened to win.
The result is distilled into an opening, middlegame, and endgame pattern library — a structured memory of motifs, plans, and conversions that the tutor can name in plain language when it teaches a student.
The methods behind the teaching
Self-Play Loop
LIORIN plays itself, scores the result, and replays the hard positions. Recursive game analysis turns every loss into a lesson and compounds strength over time.
Engine-Assisted Review
A second engine pass evaluates and validates every move. Blunders are flagged, refutations are confirmed, and only sound ideas survive into the pattern library.
Pattern Library
Openings, middlegame structures, and endgame conversions are organized as named motifs. The tutor can point to the exact pattern that applies to your position.
Move-by-Move Tutor
The LIORIN Tutor names a threat and shows a short response using the selected practice profile and current board context.
The trust loop — every move is accountable
A teaching engine is only useful if students, coaches, and districts can trust what it says. LIORIN runs every interaction through a closed feedback loop, and every cycle is written to the covenant log with a verifiable state hash. Coaches and students can also file product notes on the feedback register. When three unique voices agree on a teaching or layout theme, LIORIN applies a bounded education-safe adjustment automatically.
- 01PROPOSE
- 02VALIDATE
- 03EXPLAIN
- 04LOG
- 05REFLECT
Why the validation matters
BENCHMARK — COUNSEL-SAFE PUBLIC FRAMING
LIORIN’s public claim is deliberately simple: teaching engine, legal moves checked, accepted moves shown in the current game log. Numerical strength claims stay internal until formal testing, counsel review, and repeatable evidence support them.
The current product combines deterministic practice profiles, legal-move checks, structured teaching modes, and auditable game records. Those implemented capabilities can be tested directly; claims about playing strength or student outcomes remain withheld until repeatable evidence supports them.
- // Strength is validated by test evidence, not by marketing language.
- // In teaching modes (Trinity / Blended) LIORIN deliberately plays down to teach.
- // Public claims remain limited until benchmarking is complete.
School-safe, by design
See it in motion
Walk the full concept and potential in the pitch deck, or open the board and let the covenant log begin.