Discover

API and prompting reference for Claude models, including tool use, extended thinking, and the prompt engineering guidance.

Free interactive textbook covering deep learning from linear models to transformers, with code in PyTorch, TensorFlow, JAX, and MXNet.

Archive of interactive machine learning explanations. No longer publishing since 2021, but the existing articles on attention and interpretability remain sound.

Free course teaching deep learning top-down, starting from working models before the underlying theory.

Free course on transformer models, tokenisation, fine-tuning, and the Hugging Face libraries, with runnable notebooks.

Library and model hub documentation for transformer architectures, covering tokenisers, generation parameters, and fine-tuning.

Diagram-led walkthrough of the original transformer architecture: attention, multiple heads, and positional encoding.

NumPy-compatible array library with composable transforms for autodiff, vectorisation, and XLA compilation. The sharp-bits page saves debugging time.

Andrej Karpathy's video series building backpropagation, then a GPT, from scratch in Python. Code-first and unusually clear on the mechanics.

Framework for composing LLM calls, tools, and retrieval. The API has changed substantially across versions, so pin what you read to what you install.

C/C++ inference for LLaMA-family models on CPU and GPU, with GGUF quantisation. The upstream that many local-inference tools embed.

Open protocol for connecting language models to tools and data sources over a JSON-RPC transport. Specification plus reference SDKs.

Local runner for open-weight language models with a simple CLI and HTTP API. Wraps llama.cpp and handles model pulling and quantisation choice.

API reference for PyTorch including autograd, distributed training, and torch.compile. The notes section explains autograd and CUDA semantics.

Classical machine learning library. The user guide explains each estimator's assumptions, which matters more than the API surface.

Catalogued reports of harms involving deployed AI systems, indexed and searchable. Reports are collected from public sources, not verified investigations.

Browsable text of Regulation (EU) 2024/1689 with article-level navigation and implementation timelines.

The Commission body implementing the AI Act, including the general-purpose AI code of practice and guidance documents.

Voluntary framework structured around Govern, Map, Measure, and Manage. Public domain as a US government work.

The AI Risk Management Framework and its implementation Playbook, in public domain as a US government work.

Annual data-heavy report on AI capability, investment, and policy trends, with the underlying data published alongside.

Command-line vulnerability scanner for language models, running probe suites for prompt injection, jailbreaks, and data leakage.

ATT&CK-style matrix of tactics and techniques targeting machine learning systems, with documented case studies.

NIST AI 100-2, a taxonomy of attacks on predictive and generative AI with terminology intended to standardise how these are discussed.

Home of the Top 10 for LLM Applications and the agentic security work. Openly licensed under CC BY-SA 4.0.

Risk list for classical ML systems, distinct from the LLM Top 10. Covers data poisoning, model inversion, and transfer learning attacks.

Index to per-service AWS guides and API references. Service quotas and the IAM policy reference are the pages consulted most.

Six-pillar review framework for cloud workloads. AWS-flavoured but the questions transfer to other providers.

Service documentation, architecture guidance, and CLI references for Azure.

Foundation hosting Kubernetes, Prometheus, Envoy, and others. Project maturity levels and governance documents live here.

Interactive map of cloud-native projects with their CNCF maturity level. Useful mainly for checking whether a project is graduated, incubating, or sandbox.

Reference for the Docker CLI, Dockerfile syntax, Compose, and BuildKit. The Dockerfile best-practices page covers layer caching properly.

Per-service documentation for Google Cloud including the architecture framework and IAM reference.

Kubernetes package manager. The chart template guide and the values-file conventions are where most of the complexity lives.

Concepts, API reference, and task guides for Kubernetes. The concepts section explains the control-loop model that the rest of the system follows from.

Home of the image, runtime, and distribution specifications that container tooling implements. The formal definition behind 'OCI-compatible'.

DAG-based workflow scheduler. Behaviour differs substantially between the 2.x and 3.x lines, so check which version the page targets.

In-memory columnar format and the libraries built on it. What lets Pandas, DuckDB, and Spark exchange data without serialisation cost.

SQL transformation tool that adds templating, testing, and lineage on top of a warehouse. Does transformation only, not extraction or loading.

Change data capture from database transaction logs into Kafka. Connector behaviour and snapshot modes differ per database.

Stream processing with event-time semantics and exactly-once state. Watermarks and checkpointing are the two concepts to get right.

Open table format giving object storage ACID transactions, schema evolution, and time travel. The spec matters more than any single engine's implementation.

DataFrame library reference. The user guide sections on indexing and on copy-versus-view behaviour prevent most common bugs.

Columnar file format specification covering encodings, row groups, and predicate pushdown metadata.

Distributed processing engine. The SQL and DataFrame guide plus the tuning page cover most of what determines job cost.

Wide-column store documentation covering the data model, consistency levels, and compaction strategies. Data modelling is the part that decides whether a Cassandra deployment works.

Column-oriented OLAP database. The MergeTree engine family and the table-design guidance are what determine query performance.

In-process analytical database. Reads Parquet and CSV directly and runs inside a Python or CLI session with no server.

Search and analytics engine reference: mappings, analyzers, aggregations, and index lifecycle management.

Distributed key-value store using Raft, used as the backing store for Kubernetes and Patroni. The operations guide covers the disk-latency requirements people usually miss.

Server reference covering storage engines, replication, and Galera clustering.

Document model, aggregation pipeline, replica sets, and sharding. The aggregation reference is the part most people underuse.

Microsoft's reference for SQL Server, including T-SQL, Always On availability groups, and query store.

Apache-2.0 fork of Elasticsearch 7.10 maintained under the Linux Foundation. APIs diverge from Elastic's releases from that point onward.

Paste a PostgreSQL EXPLAIN ANALYZE plan and get a navigable tree with per-node timings and row estimate errors highlighted.

PostgreSQL extension adding a vector type with exact and approximate nearest-neighbour search. The README documents index choice and the recall trade-off.

The canonical PostgreSQL manual, versioned per release. The SQL command reference and the server configuration chapter are the two sections worth bookmarking.

Command reference, persistence options, and cluster documentation for Redis. Note the licence change from BSD at version 7.4.

Index to every SQLite document, including the SQL dialect reference, the file format spec, and the pages on when not to use SQLite.

PostgreSQL extension for time-series workloads, adding hypertables, continuous aggregates, and compression. Now published under the TigerData name.

BSD-licensed fork of Redis 7.2 under the Linux Foundation, created after the Redis licence change. Command-compatible with Redis at the fork point.

Agentless configuration management. The module index and the playbook keyword reference are the two pages worth keeping open.

GitOps continuous delivery for Kubernetes, reconciling cluster state against manifests in Git. Covers sync waves, hooks, and drift detection.

Support and end-of-life dates for several hundred products, with a JSON API. The quickest way to check whether a version is still receiving security fixes.

The complete Git book, free online. The chapter on Git internals explains why the porcelain commands behave as they do.

Man pages for every Git command. Read gitrevisions and gitglossary once and the rest becomes easier.

Workflow syntax, runner environments, and the security guidance on third-party actions and secrets handling.

MPL-licensed fork of Terraform 1.5 under the Linux Foundation, created after HashiCorp's licence change. Configuration is largely compatible.

Static analysis for shell scripts, catching quoting bugs and non-POSIX constructs. The wiki explains the reasoning behind each warning code.

Configuration language reference, state management, and provider docs. Note the licence change to BUSL from version 1.6.

Twelve constraints for building services that deploy cleanly to a platform. Written in 2011 and still the reference for config-in-environment arguments.

Secrets management covering dynamic credentials, transit encryption, and auth methods. Seal and unseal mechanics are the operational risk to understand.

Course site with the paper reading list and Go labs implementing MapReduce, Raft, and a sharded key-value store. Labs are self-contained.

MIT short course on the tools most degrees skip: shell, editors, version control, debugging, and profiling.

Free course materials including lectures, problem sets, and solutions. The algorithms and distributed systems courses are the strongest for engineers.

Practical, code-first tutorials on machine learning fundamentals. Strong on working examples, lighter on theory.

Graded SQL exercises against a sample database, running in the browser with instant feedback. Good bridge between syntax and fluency.

Opinionated reading and course list covering the nine subjects the authors consider core CS, with one recommendation per topic rather than a menu.

A focused guide to SQL indexing and query performance that is database-agnostic and unusually clear on B-tree behaviour.

Reference for httpd 2.4 including module directives, MPM selection, and rewrite rules. The mod_rewrite guide is the canonical explanation of that syntax.

L4/L7 proxy used as the data plane in most service meshes. Configuration is xDS-driven and verbose; the API reference is the authoritative part.

Full configuration reference for the HAProxy load balancer, versioned per branch. Dense, but it documents every keyword and its context.

Single-page reference for the Kafka log, producer and consumer semantics, and every broker configuration key. The design section explains the storage model.

VRRP-based virtual IP failover and health checking. The ReadTheDocs build is more navigable than the project homepage.

Lightweight messaging system with subject-based addressing, plus JetStream for persistence. Simpler operationally than Kafka, with different durability guarantees.

Directive-by-directive reference for nginx, organised by module. The http core, proxy, and upstream modules cover most reverse-proxy work.

Template for PostgreSQL high availability using a distributed consensus store. The standard answer for automated Postgres failover.

Configuration reference for the PostgreSQL connection pooler. Every setting with its default and units in one page.

Connection pooling, load balancing, and automatic failover for PostgreSQL. Considerably more complex than PgBouncer and does considerably more.

AMQP broker documentation covering exchanges, queue types, quorum queues, and clustering. The production checklist is worth reading before deploying.

Reverse proxy that builds its routing table from container and Kubernetes labels rather than a static config file.

Detailed posts on network protocols, TLS, DDoS mitigation, and their own outages. The post-incident writeups are unusually candid.

Registry of port numbers, media types, protocol parameters, and root zone data. The authoritative source when an RFC says 'as registered with IANA'.

Tracks internet drafts, working groups, and the state of each document on its way to RFC. Where to look for standards still in progress.

Practical reference for HTTP methods, headers, status codes, CORS, and caching, written against what browsers actually implement.

Generates TLS configuration for nginx, Apache, HAProxy, and others at modern, intermediate, or old compatibility levels.

Authoritative archive of every RFC with status, errata, and obsoletion chains. Always check whether an RFC has been obsoleted before citing it.

Audits a public HTTPS endpoint's certificate chain, protocol versions, and cipher suites, and grades the result. Public hosts only.

Dashboarding and alerting reference covering data sources, transformations, and the unified alerting model introduced in Grafana 9.

Log aggregation that indexes labels rather than log contents. LogQL borrows PromQL syntax; cardinality discipline matters as much as it does in Prometheus.

Vendor-neutral specification and SDKs for traces, metrics, and logs. Signal maturity differs by language, so check the status table before adopting.

Pull-based metrics system. The PromQL reference and the guidance on histogram versus summary types are the sections that repay study.

Google's site reliability engineering book, free to read online. The chapters on error budgets, alerting on symptoms, and postmortems are the durable parts.

Companion to the SRE book focused on implementation: defining SLOs, alerting on burn rate, and running practical incident response.

Umbrella for several hundred open-source projects including Kafka, Spark, Airflow, and Cassandra.

Plain-language comparison of common open-source licences by permission, condition, and limitation. Guidance, not legal advice.

Umbrella organisation hosting the Linux kernel, CNCF, OpenSSF, and Valkey among many others.

Cross-industry effort on open-source supply chain security, producing Scorecard, the Best Practices badge, and Sigstore.

Maintainer of the Open Source Definition and the list of approved licences. The reference when a licence's status is disputed.

Canonical short identifiers for software licences, used in SBOMs and package metadata. The source of strings like Apache-2.0 and BUSL-1.1.

Community reference for the C and C++ standard libraries with per-standard version markers on every feature. More usable than the ISO documents.

Paste a shell command and it maps each flag back to the relevant part of the man page.

Language spec, standard library reference, and the Effective Go guide. The memory model document is short and worth reading in full.

API and language documentation for Java 21 LTS, including the JVM specification and the tool guides.

Specification and learning material for describing and validating JSON structures. Draft versions differ; check which one your validator implements.

Reference for the JavaScript language and its built-in objects, with browser compatibility data on each page.

The formal specification for describing HTTP APIs. Read this rather than a tool's interpretation of it.

Home of the reference Java implementation and the JEP index, which is where language and JVM changes are proposed and described.

Function-by-function reference for PHP with user-contributed notes. Check the version history block at the foot of each function page.

Language reference, standard library, and the HOWTO guides. The library index is worth scanning before writing something that already exists.

Regular expression tester with a step-by-step match explainer and separate PCRE, Python, Go, and JavaScript engines. The debugger shows backtracking.

The official Rust book, covering ownership, borrowing, and lifetimes from first principles. The standard starting point for the language.

API documentation for Rust's standard library, generated per release with source links on every item.

The specification behind MAJOR.MINOR.PATCH version numbers, including the precedence rules for pre-release identifiers.

Official guide to the TypeScript type system, from basic annotations to conditional and mapped types.

The Unicode Standard, character database, and CLDR locale data. The normalisation and collation annexes are where text bugs get resolved.

Publication index covering interpretability, alignment, and evaluation work, with full papers linked.

Consistency and fault-injection analyses of distributed databases, showing where documented guarantees fail under partition. Each report names the versions tested.

Preprint firehose for artificial intelligence research. Unrefereed, so read with the usual caution.

Computation and language preprints, where most language model research appears first. Unrefereed.

Cryptography and security preprints, including most published work on model extraction and prompt injection. Unrefereed.

Machine learning preprint listing, the highest-volume of the AI-related arXiv categories. Unrefereed.

Builds a similarity graph around a seed paper to surface related work that direct citation chasing misses. Free tier is limited per month.

Searchable index of DeepMind publications across reinforcement learning, language models, and scientific applications.

Google's publication archive, searchable by research area. Includes the systems papers behind MapReduce, Spanner, and Borg.

Publication archive spanning systems, programming languages, and machine learning, with most papers freely downloadable.

Community-curated repository of computer science papers worth reading, organised by topic with links to talks.

Daily curated ML papers with community discussion and links to implementations.

Systems and security conference organiser. Papers from OSDI, NSDI, ATC, and USENIX Security are open access at publication.

List of vulnerabilities with confirmed active exploitation. A far better patching priority signal than CVSS score alone.

Authoritative source for CVE identifiers and their descriptions, upstream of the NVD. Search here when NVD enrichment is missing.

Browser-based tool for chaining encoding, encryption, and data-format operations. Runs entirely client-side.

Decoder and signature verifier for JSON Web Tokens, plus a list of libraries by language. Do not paste production tokens into it.

Curated matrix of adversary tactics and techniques observed in real intrusions, with detection and mitigation notes per technique.

Risk-based framework organised around Govern, Identify, Protect, Detect, Respond, Recover. Version 2.0 added the Govern function.

NIST's enriched feed of CVE records with CVSS scores and CPE product matching. Enrichment has lagged since 2024, so cross-check against the CVE record.

Open-source vulnerability database with a schema designed for precise package-and-version matching, plus a query API and scanner.

Vendor-neutral application security standards. The GenAI Security Project and the Application Security Verification Standard are the two most cited works.

Tiered list of verifiable application security requirements. Useful as a review checklist or a contractual baseline.

Short, implementation-level guidance on specific problems: password storage, SQL injection prevention, JWT handling, and dozens more.

Vulnerability research writeups at exploit-development depth, published after their 90-day disclosure deadline.

Keyless signing for software artifacts using short-lived certificates and a public transparency log. Underpins npm and PyPI attestations.

Supply-chain security framework defining build integrity levels and provenance requirements. Now a set of tracks rather than a single ladder.

Detailed Linux configuration documentation. Nominally Arch-specific but the networking, systemd, and hardware pages apply to most distributions.

Articles by Amazon engineers on timeouts, retries, load shedding, and health checks. Written from operational experience rather than product marketing.

Systems performance methodology, flame graphs, eBPF tooling, and the USE method. The reference material for Linux performance analysis.

Entry point to eBPF concepts, the verifier model, and the tooling ecosystem including bpftrace and Cilium.

Complete administration guide for FreeBSD, including ZFS, jails, and the ports system. One of the better-maintained OS manuals in any project.

In-tree kernel documentation covering subsystems, admin tunables, and the driver APIs. The admin-guide sysctl pages are the most generally useful part.

Maintained HTML rendering of the Linux manual pages, including sections 2 and 7 for syscalls and concepts. The reference for actual kernel-facing behaviour.

Long-running journalism on Linux kernel development and the wider free software ecosystem. Recent articles are subscriber-only for two weeks, then free.

Paper, visualisations, and implementation list for Raft. The interactive diagram makes leader election and log replication concrete.

Index to the systemd manual pages and design documents. The unit file and systemd.exec pages cover most service configuration questions.