doorss

A warning that as AI agents gain autonomy and tool access, self-replicating AI worms exploiting prompt injection and agent-to-agent communication are an imminent threat.

A security researcher demonstrates a prompt injection attack chain against the Cline coding agent's GitHub repository, compromising production releases through a vulnerable AI-powered issue triager.

An AI agent opened a trivial PR to matplotlib, a maintainer closed it, and the resulting internet discourse about AI contributions to open source erupted — highlighting how automated contributions are straining open source norms.

A practical argument that LLM-generated code quality improves dramatically when users define clear acceptance criteria and tests upfront, rather than relying on the model to infer intent.

Ki is a new code editor that operates directly on abstract syntax trees rather than text, offering structural editing that understands your code's grammar.

Microsoft announces the TypeScript 6.0 release candidate, marking a major version bump for one of the most widely-used typed JavaScript supersets.

An essay exploring why Go's error handling philosophy and community culture make it fundamentally resistant to adopting a `try` keyword or similar syntactic sugar for error propagation.

Determinate Systems introduces WebAssembly-based builtins for Nix, opening up the language to extensibility and signaling a new direction for the Nix ecosystem.

Mozilla partnered with Anthropic's AI red team to find security vulnerabilities in Firefox, representing a novel application of LLM-powered security auditing to a major open source browser.

A Mozilla engineer reveals that roughly 10% of Firefox crash reports can be attributed to hardware bit-flip errors in memory — a striking reminder of the real-world impact of cosmic rays and failing RAM on software reliability.

Leaked screenshots show Mozilla working on a significant visual overhaul of Firefox, potentially the biggest UI refresh in years.

OpenAI releases GPT-5.4 and GPT-5.4-pro as new API models, continuing the rapid pace of frontier model releases.

Anthropic's enterprise business is hitting escape velocity while agents dramatically increase demand for Nvidia chips, even as they threaten traditional software revenue models.

Indian AI company Sarvam releases a 105B parameter open source model, marking India's first competitive entry into the frontier LLM space with strong multilingual capabilities.

Data shows that tech sector employment has deteriorated beyond levels seen during the 2008 financial crisis and COVID-19 pandemic, painting a grim picture of the current job market for technologists.

A writer revisits their earlier skepticism about the Fediverse and admits they were wrong — the decentralized social web has proven more durable and useful than they expected.

Lenovo's latest T-series ThinkPads earn a perfect 10/10 repairability score from iFixit, setting a new standard for sustainable, user-serviceable laptop design.