Resource

Primary materials for technical evaluation and procurement diligence.

★ DEVICE ★ AI application Featured · primary materials + white paper Paper v10 (33 pp) / Patent claims 1-44 / 3 external benchmark data / Implementation (Rust 272 KB WASM)

SlimeTree-RLM — Measurement procedure and primary materials (semantic-driven record system)

SimpleQA / TruthfulQA / HaluEval-QA 3 external benchmarks × 3 seeds = 6,870 trials, incorrect-rate −20.4 ± 0.3 pt architectural constant. Procedure, rubrics, LLM settings, and seeds all public; access information for paper v10, patent claims 1-44, and the implementation (Rust 272 KB WASM) included.

Open the primary materials → Product page →

★ LANGUAGE ★ DEVICE cross-listed Featured · bench procedure + white paper Same-host bench vs PostgreSQL 16 / SHA-256 audit chain / SlimeCOBOL Phase D drop-in

SlimeTree-VSAM — Bench procedure and primary materials (mainframe-DB native)

Same-host bench procedure for VSAM (KSDS / ESDS / RRDS) compatible native storage vs PostgreSQL 16, measurement conditions and dataset scale, SHA-256 audit-chain alignment with regulatory requirements, and drop-in integration verification with SlimeCOBOL Phase D emit.

Open the primary materials → Product page →

Patents

SlimeCodec (JP App. 2026-046609), SlimeNENC (2026-046620), PSDP (2026-046620 / 2026-046625), SlimeTree-RLM (claims 1-44) — references will be linked here once published.

Papers

SlimeTree-RLM paper v10 (33 pp, 3 external benchmark cross-validation complete; jxiv Japanese v2 included). SlimeCodec (TCSVT submission), SlimeNENC (TSE submission planned), PSDP (paper JP v5d). PDFs posted upon acceptance.

White papers

SlimeTree-RLM primary materials (measurement procedure, LLM settings, rubrics — all public). Briefs on each of the four guarantees (domestic supply / formal guarantees / quantum resistance / 20-year support) in preparation for procurement and audit reviewers.

PoC and demos

Browser-only WASM builds (echo canceller, numerical verifier, SlimeNENC converter). Reproducible without specialized hardware.

Tools & demos

IP camera bandwidth & storage calculator covering the major codecs (H.264 / H.265 / AV1 etc.) and SlimeCodec. Live streaming demos (WebRTC / HLS switching, 4K decode, etc.).

Materials will be published progressively.

Technical deep-dives

In-depth, honestly-measured write-ups — what each tool can and cannot do, backed by real-corpus numbers (not synthetic demos). Listings are always public; some full articles require a free registration.

SlimePython — a gentle intro (turn Python into “the same-result Rust”) Public · Intro
A jargon-free entry point. Turn your Python into Rust whose output matches to the bit (proven by SHA-256). No clean rewrite needed; dynamic parts run as-is in a safe Python sandbox. From here, go deeper into the Hybrid Bit-Exact Isolate model and the PoC 40/40 results.
SlimePython PoC 0–11 deep-dive — real source, measurements, full reproduction Public · Technical
For specialists. 40/40 bit-exact; Full tier at/above python3 (the early slowdown was the shared-.so PIC → fixed by a static build); Light tier 19/20 via RustPython→WASM; auto-routing with 0 mis-routes. Real source, conversion code, measurements and a downloadable reproduction bundle (transparency is the moat).
SlimeJava — measurement procedure & primary data (turn Java into “the same-result Rust”) OPEN · technical
Full verification data for Java→Rust bit-exact. ~60,000 differential-fuzz cases (0 divergence) + 603 real OpenJDK cases, SMT proof over all inputs (20 proofs / 0 counterexamples), full-Java = Hybrid acceptance 98.7–100% (only the static core Rustified), performance (startup ~41× / RSS ~22× / flat p99), 1,053 exception-boundary crossings, seed-injection rescue. All third-party reproducible.
AS/400 (IBM i) modernization — what SlimeCL / SlimeRPG can & can’t do OPEN
CL / RPG / COBOL → portable bash & single-binary Rust. Real-corpus results: 1,711 RPG (lex 99.4%), 46 CL (Rust & bash compile 100%), 0 hangs — now verified 4/4 bit-exact on a live IBM i — plus an honest account of the compile-rate gaps.
SlimeCOBOL-ESQL — COBOL embedded SQL (EXEC SQL) conversion In development
The embedded SQL that legacy COBOL writes inside the program → Java (JDBC) / Rust (sqlx), without changing the answer. 47 cases match exactly against an independent oracle (ocesql), identical across both emitters. Final IBM DB2 validation is the next stage (currently on PostgreSQL). Not “done” — built while confirming, by the numbers, that the answer does not change.
SlimeTree-RLM — Measurement procedure and primary materials Open
3 external benchmarks (SimpleQA / TruthfulQA / HaluEval-QA) × 3 seeds = 6,870 trials, measured as a -20.4 ± 0.3 pt architectural constant. 4-LLM cross-validation shows Tier-A 8B-class LLMs converge at the 81% ceiling — the performance equalizer. Procedure, rubrics, LLM settings, and seeds all public; access information for paper v10 and patent claims 1-44 included.
SlimeNENC pipeline internals & 5-target emit→compile data REGISTRATION
Per-construct emit→compile pass-rate tables across Java / Rust / Go / Kotlin / C#, and the Slot IR design. Register to access →