AI Systems Knowledge Base — Source Index

Hyperlinks to every real system distilled in ai_systems_knowledge_base.md, grouped to match the file's Part structure.

Related context packs

File	Role
ai_fullstack_knowledge_base.md	All-in-one — systems KB + frontend design (use this when you want everything)
ai_systems_knowledge_base.md	Reasoning, agents, systems engineering; includes agent craft (Part IV = merged system architecture.md)
frontend_design_knowledge_base.md	Frontend-only slice (UI systems, spacing, motion, landings, …)
system architecture.md	Source for agent craft — merged into ai_systems_knowledge_base.md (Part IV); kept as a light standalone copy

Estimated Impact (Qualitative — Not a Measured Benchmark)

Important caveat first: no evaluation suite was run to produce these numbers — there is no SWE-bench, no held-out coding tasks, no A/B test. This is a subjective estimate of relative coding/engineering capability, scored out of 100, based on general reasoning about model tiers and about what a large, curated context document can and can't do for a smaller model. Treat it as an informed guess with stated reasoning, not as data — do not cite these numbers as measured results.

Model	Estimated score /100	Reasoning
Frontier flagship (e.g. GPT-5.x-class)	~92–96	Largest pretraining + RL investment among compared models; strongest at novel architectural judgment calls, long-horizon multi-file reasoning, and recovering from ambiguous specs without hand-holding. Context documents help these models less in relative terms — they already have deep internalized priors across most of the domains this file covers.
Claude (latest flagship)	~90–95	Comparable tier to the above; particularly strong on agentic tool-use and long-context coherence, which is exactly the category this knowledge base's Part II is about — so a curated agent-systems doc closes less of a gap here than it would for a weaker base model.
DeepSeek V4 Flash + this knowledge base as context	~78–85	A context document can supply declarative knowledge (named patterns, tradeoffs, "what to carry away" checklists) that substitutes reasonably well for prior exposure to a system's design — e.g., recognizing "this is a leaderholder pattern" or "this needs a sparse index" once it's been named and explained. It does not substitute for the model's own architecture, parameter count, or RL-trained judgment: applying a named pattern correctly to a novel, messy, partially-specified real codebase is a different skill than recognizing the pattern in a clean explanation. Expect the gain to concentrate in recall/vocabulary/breadth-of-consideration, not in raw multi-step reasoning quality.
DeepSeek V4 Flash (no added context)	~65–75	Smaller/faster-tier model; likely has real gaps in some of the more specialized domains this file covers (e.g., LSM compaction tuning, MVCC edge cases, Raft safety proofs) simply from less exposure at pretraining/RL time relative to frontier-scale models.

Why the gap between "with context" and "without context" is real but bounded: a good context document raises the floor (fewer flatly wrong claims, better vocabulary for naming what's happening, a checklist to self-check against) more than it raises the ceiling (it can't make a smaller model reason as deeply through a genuinely novel problem a frontier model hasn't seen a close analogue of). If you want an actual number instead of an estimate, the right next step is running this file as a system-prompt/context addition through a real coding benchmark (SWE-bench-style tasks) with and without it, on the same model, and diffing the pass rate — happy to help set that up if useful.

Part I — Reasoning & Math

File	System	Source
LOGIC.md	Lean 4	https://github.com/leanprover/lean4
FORMALIZATION.md	Mathlib4	https://github.com/leanprover-community/mathlib4
PROOF_SEARCH.md	AlphaGeometry	https://github.com/google-deepmind/alphageometry
PIPELINES.md	DSPy	https://github.com/stanfordnlp/dspy
DISCOVERY.md	PySR	https://github.com/MilesCranmer/PySR

Part II — Coding & Agent Systems

File	System	Source
ORCHESTRATION.md	OpenHands / Agent Canvas	https://github.com/All-Hands-AI/OpenHands
EDIT_FORMATS.md	Aider	https://github.com/Aider-AI/aider
ACI.md	SWE-agent	https://github.com/SWE-agent/SWE-agent
SDK_LAYERING.md	Cline	https://github.com/cline/cline
TOOL_SURFACE.md	Claude Code	https://docs.claude.com/en/docs/claude-code
LANGGRAPH.md	LangGraph	https://github.com/langchain-ai/langgraph
AGENTS_SDK.md	OpenAI Agents SDK	https://github.com/openai/openai-agents-python
MCP_PROTOCOL.md	Model Context Protocol	https://github.com/modelcontextprotocol/specification
LIVE_DOCS_CONTEXT.md	Context7	https://github.com/upstash/context7
12_FACTOR_AGENTS.md	12-Factor Agents	https://github.com/humanlayer/12-factor-agents
PRODUCTION_SYSTEM_PROMPTS.md	Extracted production system prompts	https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools

Part III — Systems Engineering

Distributed Systems

File	System	Source
K8S_CONTROL_PLANE.md	Kubernetes	https://github.com/kubernetes/kubernetes
CONSENSUS.md	etcd (Raft)	https://github.com/etcd-io/etcd
SHARDING.md	TiKV	https://github.com/tikv/tikv
REPLICATION.md	CockroachDB	https://github.com/cockroachdb/cockroach
DURABLE_EXECUTION.md	Temporal	https://github.com/temporalio/temporal
DISTRIBUTED_COMPUTE.md	Ray	https://github.com/ray-project/ray
LOG_AS_DATABASE.md	Apache Kafka	https://github.com/apache/kafka
LIGHTWEIGHT_MESSAGING.md	NATS	https://github.com/nats-io/nats-server

Databases

File	System	Source
QUERY_PLANNING.md	PostgreSQL	https://github.com/postgres/postgres
BTREE_STORAGE.md	SQLite	https://www.sqlite.org/src/ (docs: https://www.sqlite.org/arch.html)
IN_MEMORY_STRUCTURES.md	Redis	https://github.com/redis/redis
LSM_TREES.md	RocksDB	https://github.com/facebook/rocksdb
VECTORIZED_OLAP.md	DuckDB	https://github.com/duckdb/duckdb
EXTREME_COLUMNAR_SCANS.md	ClickHouse	https://github.com/ClickHouse/ClickHouse

Operating Systems

File	System	Source
KERNEL_SCHEDULING.md	Linux kernel (CFS/EEVDF, VFS)	https://github.com/torvalds/linux
MINIMAL_OS_DESIGN.md	xv6 (MIT)	https://github.com/mit-pdos/xv6-riscv
FROM_SCRATCH_ENGINEERING.md	SerenityOS / Ladybird	https://github.com/SerenityOS/serenity · https://github.com/LadybirdBrowser/ladybird
VERIFIED_MICROKERNEL.md	seL4	https://github.com/seL4/seL4

Compilers

File	System	Source
LLVM_IR.md	LLVM	https://github.com/llvm/llvm-project
MLIR.md	MLIR	https://github.com/llvm/llvm-project/tree/main/mlir
TREE_SITTER.md	Tree-sitter	https://github.com/tree-sitter/tree-sitter
CLANG.md	Clang	https://github.com/llvm/llvm-project/tree/main/clang

Networking

File	System	Source
ENVOY.md	Envoy	https://github.com/envoyproxy/envoy
CADDY.md	Caddy	https://github.com/caddyserver/caddy
NGINX.md	nginx	https://github.com/nginx/nginx
CURL.md	curl	https://github.com/curl/curl

ML Infrastructure

File	System	Source
VLLM.md	vLLM	https://github.com/vllm-project/vllm
LLAMA_CPP.md	llama.cpp	https://github.com/ggml-org/llama.cpp
TRITON_INFERENCE_SERVER.md	Triton Inference Server	https://github.com/triton-inference-server/server
TENSORRT_LLM.md	TensorRT-LLM	https://github.com/NVIDIA/TensorRT-LLM
SGLANG.md	SGLang	https://github.com/sgl-project/sglang
DEEPSPEED.md	DeepSpeed	https://github.com/deepspeedai/DeepSpeed

Observability

File	System	Source
PROMETHEUS.md	Prometheus	https://github.com/prometheus/prometheus
GRAFANA.md	Grafana	https://github.com/grafana/grafana
OPENTELEMETRY.md	OpenTelemetry	https://github.com/open-telemetry/opentelemetry-specification
JAEGER.md	Jaeger	https://github.com/jaegertracing/jaeger

Performance

File	System	Source
MIMALLOC.md	mimalloc	https://github.com/microsoft/mimalloc
JEMALLOC.md	jemalloc	https://github.com/jemalloc/jemalloc
FOLLY.md	Folly	https://github.com/facebook/folly
ABSEIL_CPP.md	abseil-cpp	https://github.com/abseil/abseil-cpp

Build Systems

File	System	Source
BAZEL.md	Bazel	https://github.com/bazelbuild/bazel
CMAKE.md	CMake	https://gitlab.kitware.com/cmake/cmake
BUCK2.md	Buck2	https://github.com/facebook/buck2