GitHub PR trajectories extracted by operation-prometheus for SE model training.
🔄 In a Training Loop
Raul Cardenas Montoya
rmems
·
AI & ML interests
Early-career AI/ML experimental systems and Agentic systems
https://raulcardenasmontoya.me
Recent Activity
new activity 9 days ago
rmems/eval-harness-trajectories:card: "(1 records)" -> "(1 record)" in Correction section new activity 9 days ago
rmems/Spikenaut-SNN-Telemetry:Remove CLAUDE.md (coding-agent instructions, not dataset content) new activity 9 days ago
rmems/Spikenaut-SNN-Telemetry:card: retitle zero-value section, drop "placeholder" wordingOrganizations
None yet
Synthetic data factory: SpaceXAI/Grok 4.6 🚀
Preserved for evaluation, provenance, quality analysis, and cross-generator research. No downstream model-training or weight-update
Spikenaut
My experimental spiking neural network, built from scratch: the SNN model, its supervisor trajectory + telemetry datasets, and the routing research.
Synthetic data factory: procedural and deterministic
Deterministic project-owned synthetic data: oracle simulation records and verified code-repair batches.
Synthetic data factory: Anthropic/Claude Fable 5👑
Preserved for evaluation, provenance, quality analysis, and cross-generator research. No downstream model-training or weight-update
Grok-1 dissection and ternary packing
Structural parse of open Grok-1 + route-preservation ternary GOZ1 quant experiments (Grok Build: Grok 4.5 high).
Operation Prometheus - SE Trajectories
GitHub PR trajectories extracted by operation-prometheus for SE model training.
Synthetic data factory: procedural and deterministic
Deterministic project-owned synthetic data: oracle simulation records and verified code-repair batches.
Synthetic data factory: SpaceXAI/Grok 4.6 🚀
Preserved for evaluation, provenance, quality analysis, and cross-generator research. No downstream model-training or weight-update
Synthetic data factory: Anthropic/Claude Fable 5👑
Preserved for evaluation, provenance, quality analysis, and cross-generator research. No downstream model-training or weight-update
Spikenaut
My experimental spiking neural network, built from scratch: the SNN model, its supervisor trajectory + telemetry datasets, and the routing research.
Grok-1 dissection and ternary packing
Structural parse of open Grok-1 + route-preservation ternary GOZ1 quant experiments (Grok Build: Grok 4.5 high).