Compute Tier 3 — Family Index Hub
Tier 3 family indexes are deep-dive comparative references that catalog and cross-evaluate the dominant tools, frameworks, and platforms in a specific Compute subdomain. Each index sits between Tier 2 conceptual notes (single-topic explainers) and Tier 4 implementation walkthroughs (hands-on patterns), and is structured as a comparison matrix plus selection heuristics plus 2024–26 landscape commentary.
Planned T3 family indexes
The following indexes are scoped for the Compute Tier 3 layer. Each will be ~500 lines,
frontmattered with type: compute-tier3-family, and cross-linked back into the relevant
Tier 2 conceptual notes.
ML & AI systems
- ml-framework-comparison — PyTorch / JAX / TensorFlow / Keras 3 / MLX / Mojo / Triton / Flax / Equinox / candle / tinygrad / Jittor / MindSpore + HF stack
- LLM landscape — open + closed frontier models, parameter scales, licensing, capabilities (GPT / Claude / Gemini / Llama / Mistral / Qwen / DeepSeek / Yi / Phi / Mixtral / Command-R / Reka / Grok); covers context windows, MoE vs dense, multimodal
- Vector DB taxonomy — Pinecone / Weaviate / Qdrant / Milvus / Chroma / LanceDB / pgvector / Vespa / Marqo / Turbopuffer / sqlite-vec; HNSW / IVF / DiskANN / SPANN; hybrid retrieval
Data infrastructure
- Database engine taxonomy — Postgres / MySQL / SQLite / ClickHouse / DuckDB / TimescaleDB / CockroachDB / TiDB / YugabyteDB / Spanner / Aurora / Snowflake / BigQuery / Redshift / Databricks / Iceberg / Delta / Hudi; OLTP vs OLAP vs HTAP
- Message queue + streaming — Kafka / Pulsar / NATS / RabbitMQ / Redpanda / WarpStream / Kinesis / Pub/Sub / EventBridge / SQS; durability, ordering, exactly-once semantics
Platform & ops
- Cloud provider service map — AWS / GCP / Azure / Cloudflare / Fly.io / Render / Railway / Vercel / Netlify / DigitalOcean / Hetzner / OVH; compute, storage, networking, AI
- Container orchestrators — Kubernetes / Nomad / ECS / Cloud Run / Fly / Railway / Docker Swarm / k3s / k0s / Talos; managed vs self-hosted; GitOps
- Observability tools — Datadog / New Relic / Honeycomb / Grafana stack / OpenTelemetry / Prometheus / Loki / Tempo / Jaeger / Zipkin / Elastic / Splunk / Chronosphere / Sentry
- Auth providers — Auth0 / Okta / Clerk / WorkOS / Stytch / Supabase Auth / Firebase Auth / Cognito / Azure AD / Keycloak / Ory; OIDC, SAML, SCIM, passkeys
Programming language ecosystems (ML focus)
- Programming language ecosystems for ML — Python (NumPy / SciPy / pandas / Polars / dask / Ray / modin), Rust (candle / burn / dfdx / linfa), Julia (Flux / Lux / MLJ), Swift (CoreML / S4TF legacy), Mojo, Scala (Spark MLlib / BigDL), Go (Gorgonia / Gonum), C++ (LibTorch / Eigen / Armadillo / dlib)
Conventions
- Each family index has 8–15 entries, each with: name, vendor, license, year, key strengths, key weaknesses, primary use cases, ecosystem maturity, 2024–26 status.
- Comparison axes are listed at the top (e.g. ergonomics, performance, hardware support).
- Selection heuristics translate the matrix into “if your situation is X, pick Y”.
- 2024–26 adoption trend section calls out movement (rising / stable / declining / consolidating / fragmenting).
- All entries cross-link to relevant Tier 2 notes; no orphans.
Status legend
- published
- planned
- [~] in-progress / partial