The line between human decision-making and algorithmic automation is blurring in uncomfortable ways, and a lawsuit against Meta is bringing that tension to a courtroom. Twenty-six former employees allege the company used AI-driven scoring and monitoring systems to select workers for layoffs, with the systems reportedly disproportionately targeting employees with disabilities or those who had taken protected medical leave. If proven, this would be a stark case study of a trend many are already feeling: as a popular Ask HN thread observes, software across the board is buggier, QA teams have been gutted, and the pursuit of feature velocity continues to crowd out basic reliability and human empathy. Meanwhile, the tools for building with AI are rapidly maturing, but not without their own failure modes. A fascinating piece by Martin Fowler argues that domain-specific languages (DSLs) are the key to making LLM outputs reliably useful, providing a structured harness for model outputs instead of hoping for the best from a chat prompt. This ethos is also visible in practice: Telegram quietly launched a serverless runtime for bots and mini apps, letting developers run JavaScript directly on Telegram's infrastructure without managing servers. And in a move that cements AI's transition from software to hardware, OpenAI launched the *Codex Micro*, a limited-edition desktop keypad designed to let developers monitor and control AI coding agents from their desk. In the developer tooling space, Dependabot introduced a sensible security default, now waiting three days before opening version update pull requests to mitigate supply chain attacks via malicious releases. FreeBSD 16 has completely retired all GPL-licensed code from its base system, a significant licensing milestone decades in the making. And for anyone who has ever mourned the web of the 90s, a developer painstakingly rescued over 7,000 GIFs from the ancient Ibiblio Icon Browser, preserving a piece of internet history one pixel at a time. Two bigger stories are worth noting for what they signal about the future. Google DeepMind's Demis Hassabis is calling for a U.S.-led AI standards body to review frontier models for national security risks, a proposal that could reshape how the industry self-regulates. And the biggest deal of the day is a financial one: Stripe and Advent are reportedly offering more than $53 billion to acquire PayPal. If completed, this would be a massive consolidation in payments infrastructure, bringing two of the internet's most important payment rails under the same roof.