========================================================================================
[CORE PRINCIPLES] 0% Corporate Fluff. 100% Assembly, eBPF probes, Raft & Benchmarks.
========================================================================================
- ๐ง AI Agent Runtimes & Inference: Rolling Context Compaction, Schema-Constrained Decoding (xgrammar), PagedAttention KV-cache fragmentation & SWE-bench evals.
- โก HighLoad & Distributed Storage: LSM-tree engines, Write-Ahead Logging (WAL), Raft consensus, shared-nothing cache architectures (Dragonfly/Rust).
- ๐ง Linux Kernel & Low-Level Tracing: eBPF / XDP wire-speed packet processing, io_uring zero-copy rings, CFS scheduler jitter mitigation.
- ๐ HFT & Ultra-Low Latency: Lock-free SPSC/MPMC queues, cache-line bouncing prevention, memory alignment & AVX-512 vectorization in C++23.
| Repository | Focus Area | Technology | Channel Funnel |
|---|---|---|---|
| awesome-ai-agent-infrastructure | Agent Memory Compaction & Evals | Python / Rust / vLLM | Research Post |
| awesome-highload-architecture | 1M+ RPS Distributed Engines & Raft | Go / Rust / io_uring | Join Channel |
| awesome-hft-cpp23-internals | Sub-Microsecond Lock-Free Trading | C++23 / Assembly | Read Case |
| awesome-ebpf-linux-tracing | Kernel Bypass & Observability | C / eBPF / XDP | Get Handbook |
| vllm-constrained-decoding-cookbook | Production xgrammar FSM Decoding | Python / vLLM | Architecture |
I publish unfiltered post-mortems of live production outages, assembly-level memory layouts, and architectural monographs:
- ๐ก Telegram Lab: @snowdensay โ 3 daily technical breakdowns without marketing water.
- ๐ Engineering Vault: @Snowden_book_bot โ 27 complete monographs (600k+ words, 890+ code units) with tested repositories.
- โ๏ธ Habr / VC / Dev.to: Longread breakdowns of production incidents.