Google Cloud announced Open Knowledge Format (OKF) v0.1, an open, vendor-neutral specification that formalizes the LLM-wiki pattern as a portable directory of markdown files with YAML frontmatter. OKF is not a service or platform—it requires no SDK, runtime, or registry—and a bundle renders on GitHub, ships as a tarball, or mounts on any filesystem. Each concept is one markdown file identified by its path, with only one required field (type) in the frontmatter; cross-links between files form a knowledge graph that agents can navigate without translation. Google released reference tools including a BigQuery enrichment agent, a static HTML visualizer, and sample bundles. The format targets the scattered internal knowledge problem, letting agents consume curated, version-controlled context directly, unlike retrieval-augmented generation (RAG).
A hands-on tutorial streams 3,000 documents from the FineWeb sample-10BT subset without downloading the full multi-terabyte corpus. It reproduces quality filters (Gopher, C4, custom), finding most already-passed due to pre-filtering. MinHash-based deduplication with 128 permutations and 0.7 threshold identifies few near-duplicate pairs, consistent with per-crawl deduplication. GPT-2 token counts are verified against the stored field, showing near-perfect match (mean absolute difference ~0). Analytics cover token distribution, language scores, characters per token, and top domains, providing practical insights for scaling corpus preprocessing pipelines.
Databricks released Omnigent, an Apache 2.0-licensed open-source meta-harness that standardizes the interface across terminal coding agents (Claude Code, Codex, Pi) and agent SDKs, turning them into interchangeable components. It adds a shared layer for composition (switching agents with one-line changes), contextual control (e.g., pausing at cost limits, requiring human approval for sensitive git pushes), and collaboration (sharing live agent sessions via URL). The architecture consists of a sandboxed runner with a uniform API and a policy server, and sessions sync across terminal, web UI, and mobile. An OS sandbox (Omnibox) secures credentials by injecting tokens only in approved proxy requests. Two example agents—Polly (a multi-agent coding orchestrator) and Debby (a two-headed brainstorming partner)—illustrate its patterns, and an interactive concept demo shows parallel agent delegation and policy enforcement.
This tutorial demonstrates a complete QwenPaw agent workspace setup in Google Colab. It covers installation, environment configuration, and authentication with a randomly generated password. The setup automatically configures an LLM provider from Colab secrets (OpenAI, OpenRouter, DashScope, DeepSeek, or Gemini) and creates a custom research_brief skill along with demo knowledge files. The QwenPaw Console is launched on a configurable port, made accessible via Colab proxy and optionally a Cloudflare tunnel, and a streaming chat API client is used to test the agent's responses. The result is a reproducible, secure, and extensible local-first agent platform for research and automation workflows.
On June 12, 2026, a US export control directive forced Anthropic to disable its two most capable models, Claude Fable 5 and Mythos 5, for all users because it could not filter foreign nationals in real time. The order followed a claim by another company that it had jailbroken Mythos, but Anthropic disputes this as a narrow, non-universal jailbreak. Fable 5's safety system uses classifiers that route risky queries (cybersecurity, bio-chem, distillation) to Opus 4.8 in under 5% of sessions; the model had been publicly available since June 9. All other Claude models, including Opus 4.8, remain unaffected. This appears to be the first government-forced takedown of a publicly deployed frontier AI model.
Moonshot AI released Kimi K2.7-Code, an open-weight, coding-specialized agentic model under Modified MIT license. It is a Mixture-of-Experts architecture with 1T total parameters, 32B active per token, 384 experts with 8 selected, MLA attention, SwiGLU feed-forward, and a 400M-parameter MoonViT vision encoder. The model supports a 256K-token context window, ships with native INT4 quantization, and enforces mandatory thinking mode with fixed sampling parameters (temperature 1.0, top_p 0.95, n 1). In company-reported benchmarks, K2.7-Code achieves 62.0 on Kimi Code Bench v2 (+21.8% over K2.6), 81.1 on MCP Mark Verified (beating Claude Opus 4.8’s 76.4), and demonstrates approximately 30% lower reasoning-token usage than K2.6, reducing cost and latency in agentic workflows. The 595 GB model weights are available on Hugging Face and can be self-hosted via vLLM, SGLang, or KTransformers; API access uses the kimi-k2.7-code model name with OpenAI-compatible endpoints.