4 Unconventional Ways to Cast in Typescript (8 minute read)
This article goes over four unconventional methods to cast between types in TypeScript, showing potential unsoundness in the language's type system. These methods include abusing the `is` operator, using mutation across boundaries, exploiting structural typing, and manipulating `void` type combinations. While TypeScript generally provides good type safety, these techniques can introduce subtle bugs and unsafety. | How Memory Maps (mmap) Deliver 25x Faster File Access in Go (5 minute read)
Memory maps in Unix allow a file to be mapped into virtual memory, enabling direct reads via pointers instead of slower seek/read operations. Benchmarks in Go show memory maps can be up to 25 times faster than using `ReaderAt` for random lookups and iterations. While writing to memory maps is inefficient due to page faults, they can still be great for read-heavy operations when files are cached. | | The price of mandatory code reviews (7 minute read)
Based on data from 400+ companies and 3,000+ engineers, this article challenges the mandatory code review practice in software development. The research found that teams without code reviews are 1.9x more productive but produce 2.4x more bugs, while high-quality reviews (though 38% slower) reduce bugs by 61% compared to low-quality ones. Top-performing teams (90th percentile) achieve both speed and quality by being selective about when reviews are needed. | | Workflow (Website)
Workflow is a beta TypeScript framework that simplifies building durable and reliable asynchronous JavaScript applications and AI agents. It allows devs to define workflows and steps with simple directives, handling state management and retries automatically. The framework has built-in observability with traces, logs, and metrics. | Introducing PyTorch Monarch (17 minute read)
PyTorch Monarch is a distributed programming framework that makes cluster programming easier by treating resources as local, similar to single-machine Python. Monarch organizes resources into meshes, enabling operations on entire clusters with fault handling and separation of control and data planes. It provides distributed tensors that integrate with PyTorch, allowing operations across thousands of GPUs. | Web Codegen Scorer (GitHub Repo)
Web Codegen Scorer is a tool developed by the Angular team at Google for evaluating the quality of web code generated by LLMs. It focuses on web code quality and allows users to compare models, iterate on prompts, and monitor code quality over time using built-in checks like build success, security, and coding best practices. | | Living dangerously with Claude (10 minute read)
Running coding agents like Claude with unrestricted permissions comes with increased productivity and the ability to outsource tasks, but also makes Claude vulnerable to prompt injection attacks. The safest thing to do is to sandbox coding agents, ideally on external systems with controlled network access, to mitigate these risks. | Programming With Less Than Nothing (9 minute read)
An interviewer asks a candidate to solve FizzBuzz, and the candidate responds by attempting to implement the entire problem using combinatory logic in JavaScript. The interviewer is initially skeptical, recognizing the approach as an exercise in reinventing programming primitives from lambda calculus. The candidate then constructs numbers, lists, and string manipulation functions using combinators, eventually producing the desired FizzBuzz output. | | Tyler Tries DTrace (8 minute read)
This dev explored Node.js backpressure by modifying a Node.js build to disable it, using DTrace to measure garbage collection times, and found that disabling backpressure greatly increases garbage collection effort. | When You Get to Be Smart Writing a Macro (4 minute read)
This dev improved their debugging tool, `hashp` (a better `println`), to work seamlessly within Clojure's thread-first (`->`) and thread-last (`->>`) macros by using a clever probe technique that dynamically detects the macro's context. | Surf (GitHub Repo)
Surf is an open-source, local-first AI notebook that helps users research and synthesize information from various media types by integrating files and web resources directly into a streamlined workflow. | | | Love TLDR? Tell your friends and get rewards! | | Share your referral link below with friends to get free TLDR swag! | | | |
| Track your referrals here. | | Want to advertise in TLDR? 📰 If your company is interested in reaching an audience of web developers and engineering decision makers, you may want to advertise with us.
Want to work at TLDR? 💼 Apply here or send a friend's resume to jobs@tldr.tech and get $1k if we hire them!
If you have any comments or feedback, just respond to this email! Thanks for reading, Priyam Mohanty, Jenny Xu & Ceora Ford
| | | |
0 Comments