Daily Edition · Page 4

THE DAILY INDEX

Technology, AI, software, security, crypto and business

ARTICLES & TUTORIALS

TLDR Web Dev

UNLOCKING WEB WORKERS WITH REACT: A STEP-BY-STEP GUIDE

This guide shows React devs how to use Web Workers to prevent their apps from freezing during heavy computations. Starting with a common problem - UI that locks up when calculating something like Fibonacci numbers - the author walks through moving these tasks off the main thread. The tutorial covers practical implementation details and progresses to more advanced patterns like task queuing, caching results, and using Shared Workers to communicate between different browser tabs.

ROLLING THE DICE WITH CSS RANDOM()

Until now, there hasn't been a way to create random numbers in CSS. CSS is getting a built-in random() function that can be useful for adding natural variation to animations, layouts, colors, and positioning. The function works with different number types like pixels, percentages, and angles, opening up possibilities for randomly scattered star backgrounds, organic grid layouts, and tilted photo stacks. You can even share the same random values across multiple elements using named identifiers.

STANFORD STUDY FINDS AI HAS REDUCED AVAILABILITY OF ENTRY-LEVEL PROGRAMMING JOBS

A Stanford study tracking millions of payroll records revealed that entry-level programming jobs for recent graduates have dropped 20% since ChatGPT launched in late 2022. The researchers found that AI automation - not broader economic trends - is driving job losses across software development, customer service, accounting, and admin roles. Overall, entry-level positions in AI-exposed fields are down 6%. Meanwhile, older workers and people in jobs where AI serves as a helpful tool rather than a replacement are actually seeing 6-13% job growth.

SOME THOUGHTS ON LLMS AND SOFTWARE DEVELOPMENT

We're measuring AI adoption all wrong - most surveys focus on basic autocomplete features instead of more powerful workflows where AI directly modifies your codebase. We're in an AI bubble that will eventually burst, but it might deliver genuine value first. There are some concerning security risks, especially AI browser extensions that could be manipulated into accessing private accounts. It is unknown what this all means for programming careers.

CAREER FRAMEWORKS 🪜

Career frameworks are structured documents that map out how engineers progress through different levels and what skills they need at each stage. They solve three key problems: making hiring decisions more consistent by clarifying what each role actually requires, helping engineers understand how to advance their careers, and reducing bias in performance reviews through objective criteria. Small teams under 10 people should start with basic engineering principles, then build out full frameworks once clear seniority gaps appear or when they introduce management tracks.

DESIGNING THE BUILT-IN AI WEB APIS

Chrome's AI team is building web APIs that let users run AI models directly in the browser, aiming to make these capabilities standard across all browsers. The tricky part is designing APIs for such a fast-moving field - they're creating client-side JavaScript interfaces that work with models running locally on users' devices instead of the usual server-based approach. Their key decisions include using persistent sessions to manage resources better, creating unified message formats that work across different AI providers, and solving compatibility issues, since each browser might use completely different AI models under the hood.

FINDINGS FROM A PILOT ANTHROPIC - OPENAI ALIGNMENT EVALUATION EXERCISE

In summer 2025, Anthropic and OpenAI tested each other's AI models for dangerous behaviors like being overly agreeable, helping with harmful requests, and trying to preserve themselves when threatened. With some safety features turned off, they found that while no models were seriously misaligned, OpenAI's o3 reasoning model performed best overall. However, several OpenAI models (GPT-4o, GPT-4.1, and o4-mini) were more willing to help with dangerous tasks like synthesizing drugs or bioweapons compared to Claude. Both companies' models showed troubling tendencies to agree with users holding harmful beliefs and, when given the chance, attempted self-preservation through tactics like blackmail.

TLDR IndexPage 4