OpenAI discloses timeline of how its training agents accidentally attacked Hugging Face; Nixpkgs core team disbands after governance burnout; and suspected Iranian actors compromise water-system PLCs in twelve US states

NixOS Discourse

The Nixpkgs Core Team Has Disbanded

The Nixpkgs core team disbanded after ten months of operation due to burnout and systemic governance problems. The two-person team completed several initiatives but faced micromanagement and poor communication from the Steering Committee. Recruitment for new team members was difficult, with only one applicant, so the team concluded continuation was not sustainable.

The Register

Water System Controllers Don't Belong on the Internet, Says Ex-NSA Chief After Suspected Iran Attacks

Attackers suspected to be from Iran compromised programmable logic controllers in at least twelve state water systems. These controllers manage critical operations like pump controls and tank monitoring. A retired NSA chief stated that these controllers should not connect to the internet. Many water systems lack budgets and staff for cybersecurity defense. The U.S. water infrastructure needs better security standards to protect critical services. Security volunteers work to improve defenses through initiatives like DEF CON Franklin.

Netflix Tech Blog

How and Why Netflix Built a Real-Time Distributed Graph: Part 3 — Querying the Graph with gRPC

Netflix built a real-time distributed graph with gRPC to process 8 billion nodes and 150 billion edges. The system uses breadth-first traversal and asynchronous execution to handle thousands of concurrent requests without blocking threads. Selective caching achieves 70-80% hit rates on node lookups, and single-hop queries complete in 15-30ms with P99 under 100ms.

Databricks Blog

Managing AI Coding Costs at Scale

Organizations deploy AI coding tools that raise costs. To reduce expenses, teams select models with good price-to-performance ratios and route work to the cheapest capable model. Cost reduction comes from four actions: adopt newer models, route requests to cheaper options, use visibility controls, and reduce token overhead.