Compute Library — Index

Software engineering + computer-science substance: distributed systems, databases, operating systems, networking, ML/AI fundamentals, security, observability. Sister library to Engineering/ (hardware + physics) and Languages/ (the syntactic catalog).

  • Status: Tier 1 complete (24 systems/ML deep notes + 4 theory-of-computation notes) + Tier 2 specialty (31 deep notes).
  • Library structure: mirrors Engineering / Robotics / Languages — Tier 1 deep notes + Tier 2 specialty + Tier 3 family indexes.
  • Schema: see _schema (TBD)

Subdomains

SubdomainExamples
Distributed systemsconsensus (Raft/Paxos), CAP, replication, sharding, gossip, leader election
DatabasesOLTP, OLAP, key-value, document, columnar, graph, time-series, vector
Operating systemsprocess / threads / scheduling, memory management, virtual memory, IPC, kernels
NetworkingTCP/IP, UDP, HTTP/2/3, QUIC, DNS, BGP, mesh, gRPC, WebSocket
Concurrencythreads, async, actors, CSP, locks, lock-free, transactional memory
Compilers + runtimesparsing, IR, JIT, AOT, GC, type systems
ML / AItransformers, fine-tuning, RAG, embeddings, training, inference, evaluation
Cloud / cloud-nativecontainers, K8s, serverless, IaC, service mesh, CD/CD
Observabilitymetrics (Prometheus), logs, traces (OTel), profiling
Securitycrypto, auth (OAuth/OIDC/SAML), TLS, secrets, RBAC, attack patterns
Performance engineeringprofiling, benchmarking, latency, throughput, cache hierarchy
Architecturemicroservices, monolith, event-driven, CQRS, hexagonal
Theory of computationalgorithms & data structures, complexity (P/NP), automata & computability, PL semantics

Tier 1 — Complete (24 / 24)

StatusTopicFileLinesSubdomain
Distributed systems fundamentalsdistributed-systems-fundamentals584distributed-systems
Consensus protocolsconsensus-protocols490distributed-systems
Database internalsdatabase-internals812databases
SQL vs NoSQL designsql-nosql-design564databases
OS scheduling, memory & IPCos-scheduling-memory586operating-systems
Networking foundationsnetworking-foundations605networking
HTTP/2, HTTP/3 & QUIC deephttp2-http3-quic520networking
Concurrency primitivesconcurrency-primitives588concurrency
Compiler designcompiler-design592compilers
Garbage collectiongarbage-collection540runtimes
Transformer architecturetransformer-architecture591ml-ai
Fine-tuning & RLHFfine-tuning-rlhf896ml-ai
RAG, embeddings, vector searchrag-embeddings-vector-search432ml-ai
LLM inference optimizationinference-optimization566ml-ai
Diffusion & generative-media architecturesdiffusion-and-generative-media610ml-ai
Data engineering & stream processingdata-engineering-and-stream-processing691data-engineering
Kubernetes deepkubernetes-deep595cloud-native
Containers & service meshcontainers-service-mesh593cloud-native
Observability stackobservability-stack580observability
Cryptography fundamentalscryptography-fundamentals1096security
Authentication & authorizationauth-authz772security
CPU cache & performancecpu-cache-performance506performance
Microservices patternsmicroservices-patterns552architecture

🎯 Compute Tier 1 complete: 24 / 24 deep notes, ~13,760 lines.

Tier 1 — Theory of computation (added 2026-06-03)

The classical CS-theory core, complementing the systems/ML Tier 1 above (and the mathematical foundations in Math/ — graph theory, combinatorics, logic, information theory).

TopicFileSubdomain
Algorithms & data structuresalgorithms-and-data-structuresalgorithms
Computational complexity (P/NP, the class hierarchy)computational-complexitycomplexity-theory
Automata, formal languages & computabilityautomata-formal-languages-and-computabilityautomata-computability
Programming language theory & semanticsprogramming-language-theory-and-semanticsprogramming-languages

Tier 2 — Specialty (in progress)

TopicFile
Compilers & program analysiscompilers-and-program-analysis
Type theory & type systemstype-theory-and-type-systems
Databases internals (deep)databases-internals-deep
Differential privacy & privacy techdifferential-privacy-and-privacy-tech
Distributed systems (deep)distributed systems deep
FPGA & hardware accelerationfpga-and-hardware-acceleration
iOS — Swift 6 + SwiftUI deepios-swift-and-swiftui-deep
Android — Kotlin 2 + Jetpack Compose deepandroid-kotlin-and-jetpack-compose-deep
React Native — New Architecture + Expo deepreact-native-deep
Flutter + Dart — Impeller + WASM deepflutter-and-dart-deep
Capacitor, PWA, and cross-platform webcapacitor-pwa-and-cross-platform-web
CRDTs & distributed data types (state/op-based, SEC, causal)crdts-and-distributed-data-types
CUDA + Triton + GPU programmingcuda-triton-gpu-programming
eBPF & kernel observability (XDP, kprobes, Cilium)ebpf-and-kernel-observability
Formal verification & fuzzing (model checking, SAT/SMT, AFL)formal-verification-and-fuzzing
Lock-free data structures & RDMA (atomics, DPDK, kernel-bypass)lock-free-data-structures-and-rdma
Model serving infrastructure (vLLM / TGI / Triton)model-serving-infrastructure
Model routing & LLM gatewaysmodel-routing-and-llm-gateways
Voice & realtime AI architecturevoice-and-realtime-ai
Prompt engineering & LLM agent systemsprompt-engineering-agent-systems
Agent memory & context engineeringagent-memory-and-context-engineering
LLM evaluation (benchmarks, LLM-as-judge, eval harnesses)llm-evaluation
Distributed LLM training (FSDP/ZeRO, 5D parallelism, Megatron)distributed-llm-training
Multimodal model architecturesmultimodal-architectures
Quantization & model compressionquantization-and-model-compression
MLOps & ML platform engineeringmlops-and-ml-platforms
Infrastructure as code (IaC)infrastructure-as-code
WebAssembly: component model & WASIwebassembly-and-wasi
LLM & AI application securityllm-and-ai-security
Software supply-chain securitysupply-chain-security
Confidential computing & TEEsconfidential-computing-and-tees
Randomized & approximation algorithmsrandomized-and-approximation-algorithms

🎯 Compute Tier 2: 30 / 30 done — original 10 (5 specialty + 5 mobile/cross-platform @ 2026-05-25) plus 7 systems/ML notes (CRDTs, eBPF, formal-verification + fuzzing, lock-free + RDMA @ 2026-05-18; CUDA/Triton GPU, model serving, prompt-engineering + agent systems @ 2026-05-17); plus LLM evaluation + distributed LLM training @ 2026-05-30; plus 4 additions @ 2026-05-30 (multimodal architectures, IaC, WebAssembly + WASI, software supply-chain security). (The “distributed systems deep” row above is a planned placeholder with no note yet.)

Planned (next batch): Raft implementation details, BTrees vs LSM trees (now partly covered in databases-internals-deep), FRP, TPU + accelerators.

Tier 3 — Family indexes (deferred)

  • Database engine taxonomy (Postgres, MySQL, MongoDB, Cassandra, ClickHouse, DuckDB, etc.)
  • Cloud provider service mapping (AWS / GCP / Azure)
  • ML framework comparison (PyTorch / JAX / TF / Keras)
  • LLM landscape (open / closed model catalog)
  • Container orchestrators (K8s / Nomad / ECS)
  • Observability tools (Prometheus / Grafana / Datadog / Honeycomb)
  • Auth providers (Auth0 / Clerk / WorkOS / Stytch)
  • Languages-vs-runtimes catalog

Vendor sub-libraries

Vendor-specific deep dives sit as sub-folders. Each has its own _index.md.

Sub-libraryScopeIndex
Claude product ecosystemAnthropic’s full Claude line — models (Opus 4.7 / Sonnet 4.6 / Haiku 4.5), Messages API + SDKs, Claude Code (CLI / IDE / desktop / web), Claude Agent SDK, MCP protocol, prompt caching, tool use, vision, batches, files, citations, computer use, code execution, memory, web search, hooks, subagents, skills, and the hidden-tricks compendium_index
OpenAI product ecosystemOpenAI’s full platform — models (GPT-5 / GPT-4o / o-series reasoning), Responses API + Chat Completions + SDKs, automatic prompt caching, function calling + structured outputs (strict mode), vision + audio + image generation, Realtime API (voice, WebSocket / WebRTC), Agents SDK + Codex CLI, batch + fine-tuning (SFT / DPO / RFT), and the hidden-tricks compendium_index
Google Gemini product ecosystemGoogle’s full Gemini line — models (Gemini 3.x / 2.5 / 2.0 / 1.5 Pro 2M-context), Gemini Developer API + Vertex AI + google-genai unified SDK, explicit + implicit context caching, function calling + structured output, native multimodal (text + image + audio + video + PDF), 1M / 2M long context, code execution + Google Search grounding, embeddings (multimodal gemini-embedding-2), Vertex AI Search + Model Garden, and the hidden-tricks compendium_index
AWS BedrockThe AWS-managed gateway to Anthropic Claude, Meta Llama 3/4, Mistral, Cohere, AI21, Amazon Nova / Titan, and Stability AI — full model catalog with pricing + regions, unified Converse API + tool use, Bedrock Agents (multi-agent collaboration + action groups + memory), Knowledge Bases for RAG (OpenSearch Serverless / Aurora pgvector / Pinecone / Neptune Graph + Cohere Rerank 3.5), Guardrails (content filters + PII + contextual grounding + automated reasoning), Prompt Management + Flows, cross-region inference profiles, and the hidden-tricks compendium_index
VercelThe “AI cloud” — Projects + Deployments (Preview / Production / Skew Protection / Rolling Releases / Instant Rollback), Vercel Delivery Network (126+ POPs + ISR + Image Optimization), Vercel Functions (Fluid Compute concurrency model + Edge vs Node.js + Routing Middleware), the Vercel AI SDK (Core + UI + RSC + 16+ provider plugins + AI Gateway), storage (Blob + Edge Config + Marketplace databases — Neon Postgres / Upstash Redis / etc.), Observability + Speed Insights, Cron Jobs, v0, and the hidden-tricks compendium_index
Cloudflare WorkersV8-isolate serverless on 330+ POPs — Workers runtime + APIs + Wrangler, full bindings catalog (KV / R2 / D1 / DO / Queues / Vectorize / AI / Workflows / Hyperdrive / Browser Rendering / Service Bindings), Durable Objects deep dive (SQLite-in-DO + WebSocket Hibernation + alarms), Workers AI + Vectorize + AI Gateway + Cloudflare Agents SDK, Workflows (durable execution) + Queues + Cron Triggers + Pipelines + Tail Workers, and the hidden-tricks compendium_index
Meta LlamaThe open-weight default — full lineup (Llama 2 → 3 → 3.1 405B → 3.2 vision + edge → 3.3 → Llama 4 Scout/Maverick MoE; Behemoth announced-but-unreleased), the Llama Community License (700M-MAU clause, “open weight” vs OSI-open), Llama Stack, Purple Llama safety tooling (Llama Guard / Prompt Guard / Code Shield / LlamaFirewall / CyberSecEval), deployment paths (self-host vLLM/SGLang, on-device llama.cpp/ExecuTorch/MLX, managed via Bedrock/Azure/Together/Fireworks/Groq), and the derivative-model ecosystem_index
Azure AI (OpenAI + Foundry)The enterprise model gateway — Azure OpenAI Service (GPT-5.x / o-series / image / audio) + Models-as-a-Service for Llama/Mistral/Cohere/DeepSeek/xAI/Phi via Microsoft Foundry, deployment types (Standard / Global / Data Zone / Provisioned-PTU / Batch / Developer), PTU economics, regional data residency (EU Data Boundary, Azure Government), content filtering + Responsible AI + Prompt Shields, Entra ID + Private Link + managed identity, the Foundry platform (catalog / prompt flow / evaluations / Agent Service / AI Search), and an honest Azure-vs-Bedrock-vs-Vertex comparison_index
xAI (Grok)The frontier lab with real-time X grounding — lineup (Grok-1 open-weight → 1.5 → 2 → 3 Think/DeepSearch → 4 / 4 Heavy → 4.1), OpenAI/Anthropic-compatible API + Live Search + function calling + structured outputs + vision, availability (api.x.ai / grok.com / X / Azure Foundry), benchmark positioning, and the Colossus training-cluster story_index
Open-weight providersThe build-vs-buy / self-host-vs-API economy — open-weight model families (Mistral, DeepSeek, Qwen + Gemma/Phi/OLMo) and their licensing taxonomy, the inference-provider landscape (Together / Fireworks / Groq LPU / Cerebras / DeepInfra / Replicate / Baseten + the OpenRouter & LiteLLM routing layer) compared on speed/price/quantization/data-policy, and the self-host vs managed-provider vs hyperscaler decision framework_index

Adjacent libraries

  • Engineering — physical disciplines (hardware that runs the compute)
  • Robotics — applied robotics (uses compute for perception + planning + control)
  • Languages — language syntactic + library catalog
  • Math — mathematics + statistics (underlies ML + algorithms)

How to study this library

  • Learning paths_learn_next — “if you’ve read X, learn Y next” recommendation graph plus five named multi-step tracks (Distributed-systems engineer, Database engineer, ML platform engineer, Cloud-native SRE, Security engineer).
  • Cross-cutting comparisons — two _compare_* notes that span the library: _compare_consistency_models, _compare_service-architectures.

How to use

For a software-engineering question:

  1. Identify the subdomain (distributed / DB / OS / network / concurrency / ML / cloud / security / observability / arch).
  2. Read the relevant Tier 1 deep note.
  3. Use Tier 3 family indexes to find specific tool/product options.
  4. Cross-reference Languages for syntactic / library detail.
  5. Cross-reference Math for theoretical foundations (probability, complexity, linear algebra, info theory).

73 items under this folder.