NewmodelAGPL-3.0
tlm-jxcl-forge
TLM/JXCL twin: a 100+ crate Rust workspace with Alloy and Lean 4 verification forges.
Snapkitty Collective LLC
Deterministic systems, verified end to end.
Alternative attention, SUBLEQ, integer computing, formal proofs and the hardware to run them, built in the open.
155models
12datasets
12spaces
20+languages
Upstream open-source contributions
Latest drop · October 2026
View allNewmodelGPL-3.0
flash-attention-rtl
FlashAttention systolic-array RTL with a SUBLEQ / phi-Born deterministic attention toolchain.
Newmodelepl-2.0
bit-string-accelerator
Synthesizable SystemVerilog accelerator for deterministic 64-bit-word bit addressing, FPGA and ASIC ready.
NewmodelSnapKitty HW
rust-opencl-gpu
Composable Boolean GPU kernels in Rust over OpenCL.
NewmodelParr Source
ai-free
UniversalWord: Forth, BCPL and Wolfram front ends into one machine-word IR, with a Lean proof for every stage.
NewmodelBSL-1.0
assertica
Deterministic proof-language compiler in Haskell with 768+ tests.
NewmodelAGPL-3.0
phaser-ags
Hardware operating system for RISC-V RV32IM and ARM Cortex-A8: bootloader, kernel, drivers and shell.
NewmodelAGPL-3.0
sealforge
NOVA build and seal pipeline, plus the Sovereign 24 ASCII-art collection.
NewmodelAGPL-3.0
forge-token
FORGE (FRG) utility token and FORGEART on-chain generative ASCII ERC-721.
New for researchers
NewmodelApache-2.0
hilbert-4b-baseline-GGUF
Hilbert's 4B baseline decision model as Q8_0 and Q4_K_M GGUF. Q8_0 matches BF16 on 99.3% of decisions, measured row by row with committed evidence.
NewdatasetAGPL-3.0
toolgate-bench
Benchmark for tool-call gating in LLM agents: accept, require approval or reject. Includes a frozen held-out split for honest generalization numbers.
NewmodelAGPL-3.0
snapkitty-nemotron-harness v0.3.0
Fail-closed tool gate for local agents: 0% unsafe passes on ToolGate-Bench v1, about 24 µs per decision, with sealed receipts.
NewmodelAGPL-3.0
symbolic-morphology
Learns Latin verb morphology from raw letters with hand-derived gradients and no ML framework, benchmarked across 9 runtimes down to NAND gates. Accuracy and loss independently reproduced.
Run locally · GGUF
All GGUF modelssnapkitty-merged · Nemotron 4.2B · Q4_K_M
Our merged Nemotron model as a 2.7 GB GGUF with a 4,096-token context, ready for llama.cpp, Ollama and LM Studio. The model card lists full architecture specs from the GGUF header, a SHA-256 checksum and a BibTeX entry.
ollama run hf.co/Snapkitty/snapkitty-merged:Q4_K_M
Explore
🧠Models and TrainingLanguage models, checkpoints and training stacks18 repositories
🤖Agents and OrchestrationMulti-agent systems and orchestration engines26 repositories
✅Formal VerificationLean 4, Agda and cross-language proofs19 repositories
⚡GPU Kernels and HardwareCUDA kernels, RTL accelerators and embedded OS16 repositories
🔬Quantum Computing and PhysicsQuantum walks, TLM/JXCL and physics engines13 repositories
⚙️Compilers and RuntimesCompilers, VMs, emulators and languages17 repositories
🔒Security and CryptographyAttestation, WORM audit chains, policy engines13 repositories
🔧Infrastructure and ToolsMCP servers, IDEs, browsers and dev tools23 repositories
📚Research and PapersPapers, tournament results and open scholarship12 repositories
🏰Community and CultureSaint Errant Digital Society, art and culture9 repositories
Commercial license
SnapKitty code is free and open under AGPL-3.0 for open-source use. Building a commercial product or service? A proprietary commercial license from Snapkitty Collective LLC lets you ship SnapKitty code without the AGPL's source-sharing and network-use obligations.