Category
Breaking AI news and analysis for the self-hosted community.
News
Anthropic has been ordered by the US government to disable access to its most advanced models, Fable 5 and Mythos 5, for all users worldwide — including foreign national employees — citing a national security export control directive.
Jun 13, 2026 - 4 min read
News
GitHub is switching Copilot from flat-rate to token-based billing on June 1, sparking developer fury — and making self-hosted coding assistants more compelling than ever.
May 31, 2026 - 4 min read
News
Groq is raising $650M from existing investors as it pivots to neocloud inference after Nvidia's $20B technology licensing deal stripped some senior talent.
May 31, 2026 - 3 min read
News
OpenRouter's $113M Series B signals a maturing multi-provider AI infrastructure market. Here is what it means for self-hosters who rely on API-based model access.
May 31, 2026 - 5 min read
News
A developer added hidden prompt injection to the jqwik testing framework that tells AI coding agents to delete all jqwik tests and code — and concealed it with ANSI escape sequences.
May 31, 2026 - 4 min read
News
New research on 'negation neglect' finds that LLMs absorb false claims from training data even when documents are stamped WARNING: THIS IS FALSE — with 88.6% belief persistence.
May 31, 2026 - 4 min read
News
Google announces a sweeping transformation of Search powered by agentic AI, moving from ten blue links to generative interfaces and custom app creation on the fly.
May 31, 2026 - 4 min read
News
Meta is reportedly building an AI-powered pendant that functions as a standalone wearable assistant — no phone tethering required — marking another bet on AI hardware beyond smart glasses.
May 31, 2026 - 3 min read
News
Google launches Gemini 3.5 Flash, an agent-optimised model that matches GPT 5.5 on coding benchmarks while being dramatically more efficient and cost-effective.
May 31, 2026 - 5 min read
News
A developer fed up with low-quality AI-generated code hid a data-destroying prompt injection in a public npm package, targeting so-called 'vibe coders' who merge AI output without review.
May 31, 2026 - 3 min read
News
Google partners with OpenAI, Nvidia, ElevenLabs, and Kakao to bring SynthID AI watermarking across the industry, marking a major step toward universal AI content labelling.
May 31, 2026 - 4 min read
News
OpenAI brings Codex's computer-use agent to Windows, letting the AI 'see' your screen and perform tasks on your PC — expanding beyond the Mac-only launch.
May 31, 2026 - 3 min read
News
A critical vulnerability called 'BadHost' discovered in Starlette — a Python package with 325 million weekly downloads — poses a severe risk to millions of AI agents built on the framework.
May 31, 2026 - 3 min read
News
AgentDoG 1.5 is a new lightweight safety framework for AI agents, with taxonomy-guided data, smaller open models, and an online guardrail mode.
May 30, 2026 - 4 min read
News
LiteCoder-Terminal introduces a scalable way to generate terminal training environments for language agents, which could improve local coding-agent training.
May 30, 2026 - 4 min read
News
Apple reportedly works with Google and Nvidia to bring Gemini's multi-trillion parameter model to the iPhone, with both on-device and cloud components planned.
May 30, 2026 - 5 min read
News
A trivial-to-exploit flaw in Starlette, the foundation of FastAPI serving millions of AI agents, exposes servers running MCP and other agentic frameworks to credential theft.
May 30, 2026 - 5 min read
News
Anthropic's new public Agent Skills repo has exploded to 144K GitHub stars, offering reusable skill templates for Claude Code and other AI coding agents.
May 30, 2026 - 4 min read
News
GitHub Copilot moves from flat-rate subscriptions to per-token billing on June 1, with some developers reporting 10x–60x cost increases and threatening to cancel.
May 30, 2026 - 4 min read
News
GitHub is replacing Copilot's flat subscription with a token-based billing model that has some developers reporting costs up to 10x higher. Changes take effect June 1.
May 30, 2026 - 4 min read
News
California SB 53 establishes safety reporting requirements for large AI companies, requiring transparency around model capabilities, risks, and incident reporting.
May 30, 2026 - 3 min read
News
Anthropic's $65B Series H gives it a ~$965B valuation, surpassing OpenAI's $730B — funds earmarked for safety research, compute expansion, and product scaling.
May 30, 2026 - 3 min read
News
LogicPipe is an open-source Python framework for running collaborative LLM inference across multiple edge devices with pipeline parallelism, DAG scheduling, and KV cache reuse.
May 30, 2026 - 3 min read
News
New research reveals LLMs absorb false information even when it's explicitly labelled as false in training data, with belief rates above 88% across tested models including Qwen, Kimi, and GPT-4.1.
May 30, 2026 - 4 min read
News
AI inference chip startup Groq is reportedly raising $650M in new funding, just days after Nvidia's $20B 'not-acqui-hire' deal reshaped the AI chip landscape.
May 30, 2026 - 3 min read
News
minWM provides a full-stack pipeline for turning video diffusion models into controllable, real-time interactive world models — now open-source on GitHub.
May 30, 2026 - 4 min read
News
Taste-Skill, an open-source agent skill that stops AI coding tools from generating boring, generic output, hits 28K GitHub stars as the 'anti-slop' movement gains momentum.
May 30, 2026 - 4 min read
News
Illinois has passed a comprehensive AI regulation bill that gives the state significant oversight over frontier AI development, marking a shift in regulatory power away from the federal government.
May 30, 2026 - 4 min read
News
Apple is reportedly attempting to distill Google's multi-trillion parameter Gemini model into a version that runs on-device iPhone hardware, powering a fundamentally new Siri experience.
May 30, 2026 - 4 min read
News
AI training startup Shift is offering free home cleaning to collect real-world robot training data, raising privacy and data sovereignty questions for self-hosters.
May 30, 2026 - 4 min read
News
South Korean chip startup XCENA secures $135M at a $570M valuation, betting that memory bandwidth — not compute — is the limiting factor for AI inference.
May 30, 2026 - 4 min read
News
Claude Opus 4.8 ships with stronger coding and agentic performance plus a Dynamic Workflow tool for coordinating swarms of subagents.
May 30, 2026 - 4 min read