AI

We found an undocumented bug in the Apollo 11 guidance computer code (juxt.pro) AI

A Juxt team says it uncovered an old, undocumented Apollo Guidance Computer flaw: a gyro “LGYRO” lock that is not released when the IMU is caged during a torque operation. Using an AI-assisted behavioural specification (Allium) derived from the AGC’s IMU code, they found an error path (BADEND) that would cause later gyro commands to hang, preventing realignment. The article argues this kind of resource-leak bug can be missed by code reading and emulation but surfaced by modelling resource lifecycles across all execution paths.

The Workers Opting to Retire Instead of Taking on AI (wsj.com) AI

The article examines why some workers are choosing early retirement rather than staying employed to deal with or adapt to workplace AI changes, focusing on concerns about job disruption and the burden of reskilling.

Iran threatens OpenAI's Stargate data center in Abu Dhabi (theverge.com) AI

Iran’s Islamic Revolutionary Guard Corps released a video threatening to attack US-linked energy and technology companies in the region, including OpenAI’s planned Stargate data center in Abu Dhabi, if the US targets Iran’s power plants. The report points to Stargate’s large Abu Dhabi investment and ongoing construction, while noting OpenAI has not yet responded to requests for comment. The threat comes amid broader US-Iran escalation over energy infrastructure and regional security.

Claude Is Not Your Architect. Stop Letting It Pretend (hollandtech.net) AI

The article argues that AI tools like Claude can produce plausible but context-free system designs and then short-circuit the human architecture debate, leaving teams to implement “Jenga tower” solutions they didn’t choose. It warns that architectural decisions may get rubber-stamped because AI sounds confident and “senior engineers reviewed it,” creating an accountability gap when designs fail in real production constraints. The author recommends keeping engineers responsible for design and trade-offs while using AI mainly to speed implementation.

Bernie Sanders: "AI Is a Threat to Everything the American People Hold Dear" (wsj.com) AI

In this opinion piece, Bernie Sanders argues that advances in artificial intelligence pose broad risks to workers and the democratic economy, warning that AI could undermine wages, job security, and other freedoms Americans rely on. He calls for stronger public oversight and policy action to address the potential harms from AI adoption.

Show HN: Meta-agent: self-improving agent harnesses from live traces (github.com) AI

Meta-agent is an open-source GitHub project that automates “harness” tuning for AI agents by iteratively running evaluations, collecting live traces, and generating improved harness/configuration candidates. The repository includes a quick-start workflow for running a baseline eval and then an optimization loop, plus example task definitions and configurations for tools like Claude Code. The project reports improved benchmark performance (e.g., tau-bench) and points to a WRITEUP.md for results and methodology.

Netflix Void Model: Video Object and Interaction Deletion (github.com) AI

Netflix has released VOID (Video Object and Interaction Deletion) on GitHub, an open-source pipeline built on CogVideoX that removes a target object from a video while also deleting the physical interactions the object causes (e.g., preventing a guitar from falling when the person is removed). The project includes a two-pass inpainting approach for temporal consistency, plus a mask-generation stage that uses SAM2 segmentation and a VLM (via Gemini) to produce “quadmasks” capturing both the object and interaction-affected regions. Instructions and sample data are provided, along with optional tooling to manually refine masks before running inference.