TLDR Index
TLDR Web Dev

TLDR WEB DEV 2025-07-16

5 sections14 entrieslinks blockcomplete parse

Parse notes

  • Subscription/account footer truncated before parsefooter_truncated

01ARTICLES & TUTORIALS

JAVASCRIPT SCOPE HOISTING IS BROKEN

JavaScript scope hoisting is an optimization used by bundlers to concatenate modules into a single scope instead of wrapping each in a function, but it breaks when combined with code splitting because it can't preserve the correct execution order of modules with side effects. This leads to different program behavior, with shared modules executing in the wrong order compared to unbundled code, affecting bundlers like Rollup, ESBuild, and others. The solution requires wrapping modules in functions to control execution order.

devongovett.me8 min read

WHAT'S //GO:NOSPLIT FOR?

Go's `//go:nosplit` directive removes the usual stack overflow check from functions to improve performance. Nosplit functions skip the preemption check that normally compares the stack pointer against a limit and calls `runtime.morestack` when needed, making them about 2% faster in hot loops but potentially much faster when the goroutine struct isn't cached. However, nosplit functions are dangerous because they can cause stack overflows, so the Go linker includes safety checks to make sure chains of nosplit functions don't exceed a small stack limit (792 bytes).

mcyoung.xyz8 min read

02OPINIONS & ADVICE

REFLECTIONS ON OPENAI

A co-founder of Segment, which was acquired for $3.2B by Twilio, worked for OpenAI for a year. He liked the company's bias for action, its fast decision-making pace, its focus on practical AI safety, and its commitment to broadly distributing the benefits of AI. He played a key role in launching Codex, the coding agent, within an incredibly short timeframe.

calv.info20 min read

TO BE A BETTER PROGRAMMER, WRITE LITTLE PROOFS IN YOUR HEAD

To improve as a programmer, practice writing small, mental proofs to guarantee code correctness. Strive for code with high "proof-affinity," making it easy to reason about and verify. This gets easier with consistent practice.

the-nerve-blog.ghost.io14 min read

NO TIME TO LEARN (WEB) FRAMEWORK X

Learning specific web frameworks like React may not be worth the time since they frequently change or become obsolete. Instead, focus on transferable fundamentals like vanilla JavaScript, strongly-typed languages, and understanding core specifications rather than framework-specific features like React hooks.

brainbaking.com5 min read

03LAUNCHES & TOOLS

BNA UI

BNA UI is a beautiful and customizable UI component library for React Native. It has features like light/dark theme support, Expo Router integration, TypeScript support, and zero configuration setup.

github.comGitHub repository

UZU

Uzu is a high-performance inference engine for AI models on Apple Silicon that offers a simple API, hybrid GPU/MPSGraph architecture, unified model configurations, traceable computations, and bindings for Swift and CLI.

github.comGitHub repository

REACT NATIVE APPLE LLM

The `react-native-apple-llm` plugin enables React Native apps to use Apple's on-device Foundation Models for AI tasks. It has functionalities like checking model availability, configuring LLM sessions, and generating both structured JSON outputs and natural language text.

github.comGitHub repository

04MISCELLANEOUS

WHAT CAN AGENTS ACTUALLY DO?

AI agents use LLMs to evaluate contexts, suggest and utilize tools to enrich those contexts, and manage flow control for tool usage. Customer support and bug triaging are good use case examples, and AI agents should be treated as software products requiring careful design, iteration, and human oversight.

lethain.com17 min read

ASYMMETRY OF VERIFICATION AND VERIFIER'S LAW

"Asymmetry of verification" is where some tasks are much easier to verify than to solve, which is an important concept in improving AI. Examples include Sudoku, website operation, and BrowseComp problems, where easy verification enables AI training. The "Verifier's Law" states that AI's ability to solve a task is proportional to its verifiability.

jasonwei.net6 min read