Update README.md

docs: clarify Tier 1 fast-path training diagnostics vs Tier 2 peak full-brain inference in metrics_summary.json

docs: clarify Tier 1 fast-path training diagnostics vs Tier 2 peak full-brain inference in technical_report.md

docs: clarify Tier 1 fast-path training diagnostics vs Tier 2 peak full-brain inference in README.md

fix(metrics): clarify single-GPU execution on 2x T4 node & in-memory Triton JIT compilation

fix(inference): add automatic relative script_dir path resolution for weights and vocab

docs: clarify verified synaptic footprint table in technical report

docs: clarify checkpoint byte breakdown with explicit subtotals for inference weights and training traces

Load spatial coordinates directly from checkpoint buffer in pure shell benchmark

Load spatial coordinates directly from checkpoint buffer; clean rotary layer notation

Update custom benchmarks master executive report with current checkpoint SHA-256

Sanitize rotary notation in standalone reproduction script

Update technical report with sanitized coordinate buffer references

Update documentation with theoretical formulations and sanitized architecture

Sanitize rotary layer variable names and load spatial coordinates from checkpoint buffer

Bake spatial positional embeddings (pos_coords) directly into model buffer (SHA-256: ca1a4ae2)

Purge __pycache__ bytecode from custom_benchmark_answer_sheets

Add standalone reproduction script recreate.py and verified Q&A audit sheets

Add 3 Custom Silicon Benchmarks to benchmarks/README.md and redirect to custom_benchmark_answer_sheets

Update 01_niah_128k_answer_sheet.md with Verifiable Q&A Audit Log

Update README.md with 3 Custom Silicon Benchmarks (128K NIAH, Continual Learning, Lyapunov Attractor)

Add 3 Custom Silicon Benchmark results: 128K NIAH, Continual Plasticity, Lyapunov Attractor Audit

Update benchmarks README: 12k-15k internal thinking steps/sec and silicon kernel speeds

Update throughput audit: 12k-15k internal thinking steps/sec and silicon kernel speeds

Update Quickstart to Full Cognitive Dual-Engine Brain Inference (Peak Condition)

Add official standalone full dual-engine brain inference.py in peak condition

Add NVIDIA Tesla T4 Hardware Specs and Exact Benchmark Inference Speed (TPS) Audit

Add NVIDIA Tesla T4 Hardware Specs and Exact Benchmark Inference Speed (TPS) Audit

Update README with explicit cognitive brain improvement breakdown & benchmark_answer_sheets paths

Update benchmarks with explicit cognitive brain improvement analysis

Add Official 15-Suite Benchmark Answer Sheets (Pure Shell and Full Model)

Remove obsolete answer_sheets folder in favor of benchmark_answer_sheets

Update README with Official 15-Suite Two-Tier Academic Benchmark Leaderboard

Add Official 15-Suite Benchmarks (Runners, Metrics, and Logs for Pure Shell and Full Model)

Add Official 15-Suite Benchmark Answer Sheets (Pure Shell and Full Model)

Add full-precision accuracy benchmark comparison and 131K stream architecture pillars

Add full-precision accuracy benchmark comparison and 131K stream architecture pillars

Update README.md

Headline 100% custom Triton megakernel training engine in technical report

Headline 100% custom Triton megakernel training engine (Zero PyTorch autograd)

Update tags in README.md

Official release of Synaptic Edge 10M 1-Billion Token 3D Topological Silicon Model

initial commit