AI-Generated Code Found in rsync, Backup Breakage Sparks Controversy
A bug was reported in version 3.4.3 of **rsync**, the file synchronization tool widely used across UNIX/Linux environments, which broke incremental backups. When a user dug into the commit history, they discovered that dozens of commits since version 3.4.1 had been recorded under the joint byline "tridge and claude" — rsync developer Andrew Tridgell and Claude, Anthropic's AI assistant.
The discovery prompted a strongly worded protest post on GitHub titled "Please Do Not Vibe Fuck Up This Software," and the debate spread to Reddit and Hacker News, turning into a reckoning over whether AI-generated code belongs in critical open source infrastructure.
Tridgell pushed back in a Medium post. He explained that he used AI only for the "heavy lifting" of "migrating the test suite's old shell scripts to Python," that he did the design himself, and that he manually reviewed the generated code. "I am a software engineer with 40 years of experience," he argued. "I did not simply vibe code 'convert the test suite to Python.'"
The episode also highlighted the growing maintenance burden on maintainers flooded with AI-generated security reports, throwing into sharp relief the challenges that open source maintenance faces in the age of AI.
US Department of Defense Operates AI Propaganda Site Aimed at Latin America
An investigation by The Intercept revealed that the US Department of Defense runs **La Tilde**, a heavily AI-driven propaganda site targeting internet users in Latin American countries.
La Tilde publishes articles in both Spanish and English under the tagline "putting an accent on the news." Alongside personal finance articles, the site mixes in content portraying US military operations in a positive light, including pieces praising the operation to apprehend Venezuelan President Maduro.
The site carries no bylines or listed editorial staff, and when analyzed with AI-text detection tools, multiple articles were flagged as machine-generated. Emerson Brooking of the Atlantic Council described the site as "saturated with AI through and through," noting that despite the low quality, the ability to "generate new content and news sites with the flip of a switch" could dramatically change the speed and scale of propaganda.
La Tilde is operated as a military messaging platform for US Special Operations Command South (SOCSOUTH), and it also emerged that the web design was subcontracted to Antpack, a Colombian digital marketing company. Country-specific versions for seven nations — including Ecuador, El Salvador, Honduras, and Panama — are already in the works.
Walmart Shareholders Vote Down Proposal for AI Workplace Impact Report
According to Reuters, a shareholder proposal calling for a report on the impact of AI use on employees was rejected at Walmart's annual shareholder meeting.
The proposal was submitted by the investor group United for Respect. Ava Williams, a Walmart overnight employee in Spokane, Washington, testified that AI-driven work standards impose "impossible timelines" and cause "injuries, burnout, and high turnover." They also pointed to pressure to skip safety procedures such as sanitizing and checking for expired products.
Walmart said it has an AI philosophy emphasizing "responsible use and human judgment." The company is currently pushing automation across its warehouses and stores, with more than 60% of stores receiving goods from automated distribution centers and more than 50% of e-commerce orders handled through automation.
Walmart is the largest private employer in the United States, employing roughly 1.6 million people. As AI adoption accelerates, the vote has once again raised the question of who monitors the impact on workers.
agentgateway Joins Linux Foundation's AAIF — An Open Gateway for AI Agents
**agentgateway**, an open source gateway for AI agents, has joined the Agentic AI Foundation (AAIF) under the Linux Foundation as its fourth hosted project.
agentgateway is a unified gateway that manages MCP (Model Context Protocol) traffic, agent-to-agent communication (A2A), LLM inference, REST APIs, and gRPC services through a single operational surface. It is built in Rust and runs with high throughput and low latency.
Key features include JWT authentication, RBAC, mTLS, prompt guards, rate limiting, and content-based routing, with support for Kubernetes and bare-metal environments. More than 300 contributors from over 60 organizations — including Red Hat, Adobe, Salesforce, and Microsoft — are already involved. It is available under the Apache 2.0 license.
proveKV — An Open Source Tool Achieving 36x Lossless KV Cache Compression
A technique called **proveKV**, which losslessly compresses LLM KV caches by up to 36x (compared to f32), was announced on Reddit's LocalLLaMA community.
Validated on SmolLM2-1.7B with WikiText-2, it achieves 36x lossless compression and 68x lossy compression with no PPL (perplexity) degradation. Claims are managed declaratively as CLAIMS.json and can be verified with an automated audit script (prove_audit.sh). The Rust implementation is available on GitHub.