RC RANDOM CHAOS

The Wire

Curated cybersecurity and tech news — AI-summarized, source attributed.

open-sourcetech-culture

Free Universal Construction Kit: 3D-Printed Adapters Bridge 10 Toy Systems

F.A.T. Lab and Sy-Lab released a set of nearly 80 two-way adapter bricks that let ten otherwise incompatible children's construction toys interconnect, includin

via Hacker News ·
aitech-culture

Goodhart's Law at Scale: How LLMs Broke the Proxy Measures of Knowledge Work

Knowledge work has always been hard to evaluate directly, so reviewers fall back on proxy signals — clean prose, polished formatting, code that reads cleanly on

via Hacker News ·
aicloud

Google commits $10B to Anthropic now, $30B more on milestones at $350B valuation

Alphabet is deepening its entanglement with Anthropic through a staged capital injection: $10 billion in cash up front at a $350 billion valuation — flat to Feb

via Hacker News ·
tech-cultureopen-source

Hokusai's Tessellation Studies: Geometric Patterns from a Master Printmaker

A digitized work from Japan's National Diet Library showcases tessellation studies attributed to Katsushika Hokusai, the Edo-period artist best known for The Gr

via Hacker News ·
open-sourcetech-culture

Martin Galway open-sources his 1980s Commodore 64 game music source

Legendary SID composer Martin Galway has published the original source files for his Commodore 64 game music on GitHub, covering work from the mid-1980s includi

via Hacker News ·
tech-culturedevops

Microsoft collapses Windows Insider channels to two, kills gradual rollouts in Beta

Microsoft is restructuring the Windows Insider Program into two channels, Experimental and Beta, replacing the current Dev/Canary/Beta split that the company co

via BleepingComputer ·
aidevops

OpenAI folds Codex into main GPT-5.5 line, ends separate coding model

Romain Huet confirmed that OpenAI has discontinued the separate Codex model branch as of GPT-5.4, merging coding-specialized capabilities into the unified flags

via Simon Willison ·
aiprivacy

OpenAI rolls out Privacy Filter to scrub sensitive data from model interactions

OpenAI announced a Privacy Filter feature designed to detect and redact personally identifiable information and other sensitive data before it reaches model tra

via Hacker News ·
aidevops

OpenAI tells devs: don't port GPT-5.4 prompts to GPT-5.5, rebuild from scratch

OpenAI shipped GPT-5.5 in the API alongside a prompting guide that explicitly warns against treating it as a drop-in replacement for gpt-5.2 or gpt-5.4. The gui

via Simon Willison ·
tech-cultureai

Overthinking as sabotage: when scope creep kills personal projects

Kevin Lynagh argues that personal projects fall into two camps: ones he just builds and finishes, and ones he sabotages by surveying prior art, expanding scope,

via Hacker News ·
tech-culturepolicy

Palantir staff break internal silence over ICE work and Trump-era contracts

Current and former Palantir employees are openly questioning the company's trajectory under the second Trump administration, with internal conversations now fra

via Ars Technica ·
malwarecybersecurity

Pre-Stuxnet 'fast16' malware surfaces, targeted engineering software years before 2010

Researchers have identified a previously undocumented malware family dubbed 'fast16' that predates Stuxnet and was aimed at engineering software environments. T

via The Hacker News ·