Awesome notes

When I found an interesting article or resource, if something opened my eyes I documented it here.

*The AI-Native Software Development Lifecycle - White paper proposes the emergence of a fully AI-native SDLC, where AI is integrated seamlessly into every phase of development, from planning to deployment.

• Temporal Logic of Actions - TLA is a logic for specifying and reasoning about concurrent and reactive systems by Leslie Lamport.

• How to Write a Git Commit Message - commit messages matter, how to write them well.

• Chinchilla - how to estimate optimal model size and number of tokens for training a transformer language model under a given compute budget.

• WebAssebly Java runtime - run native code without JNI in a portable way using Endive.

• SEDA - staged event-driven architecture.

• Stop Using Pull Requests - pull requests are an antipattern for private software teams.

• JVM Anatomy Quarks - mini-post series, where every post is describing some elementary piece of knowledge about JVM internals by Aleksey Shipilëv.

• Spaced Repetition - Learning technique with flashcards to align to forgetting curves.

• Email Addresses - Deep dive into email addresses, a lot more than a username and a domain.

• Java Compact Header Layout - Java 25 (JEP 519) to reduce the size of object headers in the Hotspot JVM on 64-bit architectures.

• Java Evolved - A collection of modern Java code snippets. Every old Java pattern next to its clean, modern replacement — side by side.

• Java Almanac - Systematic collection of information about the history and the future of Java.

• The Power of Data Sketches - summary of Theta Sketches, HyperLogLog, Compressed Probabilistic Counting, HyperLogLog Map, Tuple Sketches, Classic Quantiles Sketches, KLL Sketches, REQ Sketches, Frequent Items Sketches and Frequent Distinct Tuples Sketch.

• GC: From First Principles to Modern Collectors in Java, Go and Python - GC recap from McCarthy (1960) and Wilson (1992) papers to Java ZGC.

• SQL Differential Privacy - Add some noise to query result to protect individual privacy.

• How Google SREs Use Gemini CLI to Solve Real-World Outages - Incident handling at Google.

• Mean time to mitigation - Generic mitigations (rollback, degrade, upsize, block, drain, quarantine) to shorten mean time to mitigation during outage.

• BigQuery under the hood - Colossus (filesystem), Dremel (query engine), Borg (compute) and Jupiter (network).

• A Deep Dive into Common Open Formats for Analytical DBMSs - Parquet, ORC, Arrow, Arrow/Feather.

• Google Zanzibar - Google system for storing and evaluating ACLs.

• Fluent Python - Companion website for the lizard book.

• The Zen of Python - beatiful design rules.

• Beyond Indexes: How Open Table Formats Optimize Query Performance - Compares traditional relational database indexing with open table formats like Iceberg, Delta, and Hudi, explaining why analytical systems rely on data layout rather than B-tree indexes for performance.

• Smart Home LLM leaderboard - LLM models evaluation for Home Assistant assists, questions and automations.

• How I Scanned all of GitHub’s “Oops Commits” for Leaked Secrets - Github never forgets commits.

• 25 self hosted services - Ultimate guide for 25 must have home server services.

• Vision language models are blind - VLMs are surprisingly still struggling with low-level vision tasks that are easy to humans.

• Git configuration - How core git developers configure git.

• BIRD-SQL - BIg Bench for LaRge-scale Database Grounded Text-to-SQL Evaluation, over 12,751 unique question-SQL pairs, 95 big databases. It also covers more than 37 professional domains.

• Inner Source Patters - how to use of open source principles and practices for software development within the confines of an organization.

• Cost optimization of Dataflow pipelines - beyond optimization, a collection of references to valuable materials.

• How we built a new powerful JSON data type for ClickHouse - variant and dynamic types, dense storage, arbitrary nesting.

• Consistency Models - consistency models for concurrent systems.

• Big Data is dead - by Jordan Tigani, ex googler, DuckDB advocate.

• Apache Druid tuning guide - how to achieve sub-second responses.

• Apache Beam Design Documents - These documents, managed by Googlers, provide an explanation of the Beam programming model and more.

• Dataflow Streaming Execution Model - Processing within exactly-once guarantees and with latencies consistently in the low single-digit seconds.

• How NAT traversal works - How to traverse firewalls, SNAT, CGNAT in mesh networks.

• Tracking health over debt - How to evaluate projects quality using lenses for: development, operations and architecture.

• How to write efficient Flink SQL - What is a difference between: dual-stream, lookup, interval, temporal and window joins.

• Part 1, part 2 and part 3 - Fleet-wide refactoring at Spotify, how to get 70% adoption of shared library in 7 days?

• Lessons from Building Static Analysis Tools at Google - Finding bugs is easy, but how make them actionable at scale?

• BigQuery metadata - When Metadata is Big Data.

• Prompt engineering - How to talk to GenAI (in polish)?

• Datadog incident - Multi-region and multi-cloud architecture didn’t necessarily help to avoid the outage if you forget about legacy security in-place update channel.

• The Engineer/Manager Pendulum - You don’t have to choose to be one or the other, but that you must choose one at a time.

• JavaScript event loop and Node.js event loop - Runtime models for JavaScript in the browser and server side.

• Learn You a Haskell for Great Good! and Learning Scalaz - Side by side comparision.

• Valhalla - Mechanical sympathy for Java.

• test && commit || revert - How to reduce size of the change to increase velocity? Revert the change always if tests fail!

• Learning different technologies based in the Scala Programming Language - Coding exercises for Standard library, FP in Scala, Cats and other libraries.

• Abstract Type Members versus Generic Type Parameters in Scala - trait Collection[T] {} vs. trait Collection { type T }.

• The Principles of the Flix Programming Language - Design principles of Flix programming language and Flix compiler. Awesome checklist.

• Local-first applications with automerge - Conflict-Free Replicated Data Type (CRDT), binary protocol, lossless snapshots, JavaScript and Rust (WebAssembly) implementations.

• Re-architecting Apache Kafka for the cloud - Tiered storage, Kraft cluster management (instead of ZK), multi-tenancy (consumer/producer quota, threads, connections, replication quota).

• Understanding the Python GIL - Global interpreter lock, the reason why Python doesn’t scale on multi-core architectures.

• Roblox Return to Service 10/28-10/31 2021 - Postmortem from non-trivial 73h outage due to Consul failure, why it took so long? Consul monitoring didn’t work without Consul 😜

• Exception Handling Considered Harmful - If you have always perceived that exception handling doesn’t really work.

• The Scheduler Saga - Implement Goroutines from scratch: global run-queue, distributed run-queue, work stealing, handoff, cooperative preemption, FIFO vs LIFO.

• A Guide to the Go Garbage Collector - Interesting how it differs from JVM, one or two knobs for configuration.

• Minimal version selection (MVS) - Dependency management algorithm to produce high fidelity builds.

• Documenting architecture decisions - One or two pagers with the motivations behind important architecture decisions, without ADRs this kind of knowledge evaporates.

• Code review automated comments - Framework used by many tools for commenting code reviews (documentation linters, static code analysis, test coverage).

• Write a documentation like a pro - Apply Microsoft or Google style guide for your writing.

• Test sizes - Google way for tests classification: small, medium, large (in addition to test scopes: unit, integration, e2e).

• Pre-commit: Don’t Git hooked! - Don’t use Git pre-commit hook for checking code quality, valuable checks take too much time to run.

• Don’t deploy applications with Terraform - Terraform is for static resources management, not for application deployment.