Parse notes
Canva developed a system that used generative AI to create a synthetic dataset of realistic queries and designs. This let the company perform offline evaluation of search algorithm changes. The system was meant for privacy, which meant that user queries were hidden from training. The approach allowed for rapid, reproducible testing of code changes locally, speeding up Canva's development process.
Slack's log search engine, Astra, initially used fixed-size data chunks, leading to wasted disk space and increased infrastructure costs due to underutilized chunks. To address this, the Slack team redesigned Astra to use dynamic chunks. This involved implementing a first-fit bin packing algorithm to optimize chunk assignment across cache nodes, using Zookeeper for persistent data tracking.
The problem with legacy code isn't the code itself, but rather poor knowledge transfer and maintainability among developers. Long programmer tenure, well-architected and documented code, effective knowledge-sharing practices, and the use of standardized tech and frameworks can prevent code from becoming legacy code.
A "codin' dirty" approach is better than following a "clean code" philosophy. This means writing larger, contextually relevant functions, with integration tests over unit tests for their stability and higher-level perspective. It also means minimizing the number of classes and interfaces to simplify project complexity.
Complex code, persistent frustration with debugging, and the rise of AI tools can make developers fear that they are not good enough. Progress, even if imperfect, constitutes "good enough.”
SQLiteStudio is a free, open-source, cross-platform application for creating, editing, and browsing SQLite databases.
Rill is a lightweight Go toolkit that makes concurrent programming easier by providing composable, channel-based functions for common concurrency tasks, like batching, error handling, and order preservation.
A rendering of over a million of the web's top homepages.
Employee performance at Fortune 500 companies is probably best represented by a Pareto distribution (rather than the common Gaussian distribution). This Pareto distribution implies that the bottom 10% of performers are not drastically different from the next 10%, negating the justification for annual firings based on a forced curve.
Apple's system aggressively prioritizes tasks based on Quality of Service (QoS) classes (User Interactive, User Initiated, Legacy, Utility, and Background). This influences CPU scheduling and I/O. To avoid priority inversion bugs, use specific synchronization primitives like `os_unfair_lock` or `pthread_mutex`, which enable QoS inheritance, boosting lower-priority threads when accessed by higher-priority ones.
A new research paper proposes treating LLMs as "Method Actors" to improve prompt engineering. This approach frames prompts as scripts, guiding the LLM through a performance, similar to how a director guides an actor. The "Method Actors" model breaks down complex tasks into smaller, manageable subtasks, leading to better performance on certain tasks.
GitHub Actions has announced several upcoming breaking changes, including a migration of the `ubuntu-latest` label, the shutdown of Artifact actions v3, updated workflow validation for pull requests from forked repositories, a new webhook rate limit, and network allow list updates for self-hosted runners.
A native, performant bottom sheet component for React Native apps, with features like blur support and asynchronous ref methods.
Tailscale's small infrastructure team (only three engineers) manages a large-scale network efficiently by using its own product.
Amazon S3 now supports conditional writes, allowing users to prevent accidental overwrites of objects by verifying the object's ETag before updating it.