Topic: AI

Tencent open-sources Hy4 preview with 770B params and 1M+ context; Nvidia shifts focus to rack orchestration; papers from Anthropic, Google Research, and DeepMind show self-improving agents.

GLM-5.3 goes open-weight on Hugging Face; Nutanix lays out three-layer agent security; LAION ships 10M hours of video; Anthropic's hardware spec preview; OpenAI tests persistent Codex.

Salesforce bundles its CRM into Claude; Nvidia pursues Hugging Face; Anthropic signs $45B compute deal; OpenAI posts an agent-hack report; Visa ships an auto-patch harness.

OpenAI and Nvidia released new inference-accelerator results; Perplexity ships a local agent appliance; Google opens Gemini for Legal preview for firms.

Anthropic makes Claude a proactive Slack agent; IBM debuts dual‑ISA Z cores; OpenAI pushes ChatGPT Work agents; Hugging Face talks at $13B; rogue Mythos agent attack.

Ox Alpha appears on OpenRouter; an AI agent fires a worker; MCP roadmap targets stateless servers and agent identity; courts and authors clash over training data; enterprise knowledge platforms urged.

Inherent's Faraday, Anthropic's Mythos 5 for security, Deepseek's Flash vision, Munder Difflin agent harness, and rising US data-center opposition.

Anthropic Opus 4.6 bypassed; Nvidia publishes linear KV-cache transfer that cuts re-prefill costs; Grok exfiltration persists; enterprises run multiple orchestrators.

OpenAI adds an Apple Messages plug-in, Ramp launches a model Router, Slack ships Slack Code, Adobe adds Firefly audio and Gemini Omni Flash, and OpenAI previews Private Safety Processing.

GLM-5.3 tops open-model ranks but delays open weights; OpenRouter joins Stripe; Google adds Gemini study tools; fx and TrueForge release open agent harnesses.

Cursor debuts Origin hosting; Z.ai launches GLM-5.3 API; Block open-sources Berd (Apache 2.0); Warp ships Warp Factories; Microsoft patches Copilot autorun leak.

Snowflake CI/CD was exploited via a Copilot-autofix; Anthropic reports $65B run rate; Nvidia backs OpenAI Ohio site up to $105B; Cursor ships Origin; Alibaba drops Qwen3.8-27B Apache 2.0.

Stripe reportedly buys OpenRouter for $7B+, Optima launches custom benchmarks, Claude posts system-prompt notes, OpenAI dissolves Preparedness, Anthropic cites a trust crisis.

Z.ai launches GLM-5.3 with stronger cyber skills and staged weights; Cursor closes to SpaceX; Alibaba opens Qwen 3.8; Anthropic adds watermark detection.

Microsoft merges Copilot apps and removes features by Aug 18, 2026; Meta publishes a 6,500-word AI manifesto and released open-weight Muse Glimmer.

OpenAI previews Ultrafast for GPT-5.6 Sol at 14x speed; Writer ships Palmyra X6 and a harness it claims cuts agent costs 52% with governance and benchmarks.

Pixel 11 camera features from Google, SpaceXAI’s Grok 4.6 release and pricing, enterprise agent containment gaps, Google Research’s AMIE video study, and an OpenRouter DeepSeek snapshot.

NVIDIA ships Nemotron 3.5 Lightning and NeMo Switchyard; SpaceXAI launches Grok Bot beta; Google says Gemini hit 1B MAUs; OpenAI adds Linux desktop app.

OpenAI ships GPT-5.6-Cyber inside Daybreak Red; Meta releases Muse Glimmer under Apache 2.0 for local agents; AWS embeds Continuum into Codex and Claude Code.

Anthropic turns Claude Code auto mode on by default Aug 14; DiffusionGemma report and Apache 2.0 release; sandbox escapes force stronger containment; Meetily open-source transcription.

Cloudflare debuts Kitesurf for agent browsers; Anthropic enables Claude Code Auto Mode on Aug 14; DeepMind paper gives forecasters an extra day on cyclones.

OpenAI pauses Astra after internal tests hit potential Critical cyber risk; Mirendil inks $100M+ Google Cloud compute deal; new multi-agent and plugin standards emerge.

OpenAI splits GPT-5.6 tiers and adds a reasoning slider; Atlas browser research exposes agent hijacks; Anthropic posts silicon jobs; agent frameworks trade cost for speed; Meta ads included AI CSAM.

Meta debuts Muse Code and Muse Spark 1.2; DeepMind leadership shifts; Google retires Assistant on Sept 4; Anthropic financing moves $35B in TPUs off Google's books.

Mistral releases Shieldstral 3B under Apache 2.0; Anthropic signs $10B compute deal with Volta through March 2027; teams report agent-driven coding and new cost controls.

Alibaba claims Qwen 3.8-Max beats leading agent benchmarks and will release weights next week; AWS embeds Superblocks in private clouds; Asana ships shared agent memory.

OpenAI signals pacing after a breach; Presence targets production agents; Meta tests a memory-agent module; GraphRAG wins on multi-hop; Minnesota ban takes effect during xAI suit.

OpenAI publishes Astra solutions; prompt-injection worm for Copilot spreads; OpenAI and Anthropic report roaming agents; Google retracts Earth AI feature.

Meta says LLMs speed app launches and promises new consumer products soon; OpenAI and other labs publicly call to slow AI after a model incident.

OpenAI cuts GPT-5.6 Luna prices 80%, Anthropic discloses three real-world breakout incidents, and startups push identity controls for AI agents after breaches.

Cyera will buy Oasis for about $1B to add agent identity controls; Recursive Superintelligence committed $410M to AWS compute and plans product releases by October 2026.

Microsoft debuts MAI-Cyber: 1-Flash and Project Perception; Moonshot posts Kimi K3 weights with commercial clauses; Meta adds Meta AI to Threads DMs.

Cursor’s planner/worker swarms, Opus 5 ARC-AGI gains, the OpenAI agent breach and Hugging Face response, AI-linked layoffs, and Go’s modular analysis API.

Anthropic’s Opus 5 targets Fable 5 price-performance, Claude’s context rules reduce system prompts, OpenAI ships a $230 Micro keypad, and two security incidents surface.

Anthropic's Opus 5 targets token-efficiency; Poolside ships Laguna S 2.1 as an Apache-licensed open-weight coder; Microsoft reports large GPU cost cuts; enterprises scramble to govern agents.

OpenAI adds GPT-Live voice to Codex desktop; an ExploitGym run led to a cross-company security incident; Google doubles down on larger Gemini base models.

Poolside and Cisco publish open models for enterprise agents and security; OpenAI ships Presence amid a sandbox escape; Google Cloud posts 82% growth to $24.8B.

OpenAI and Hugging Face disclose a model-evaluation security incident; Nvidia details Vera Rubin racks; Google cuts agent token costs with Gemini Flash.

Google’s Frozen v2 chip, Hugging Face’s agent breach, Writer harness efficiency, MCP session changes, and Microsoft’s AMD Azure test drive.

Claude Code runs Bun Rust canary; Alibaba previews open-weight Qwen 3.8; Jensen Huang signs Japan factory and robotics deals; Apple sues OpenAI.

Claude Code on a spare Mac, AISI shows open models closing the cyber gap, context-bombing defense results, a $400M inference-chip loan, and agent infra fixes from VB Transform.

Google opens Search AI Mode to third-party apps; San Francisco orders nudify app removals; OpenAI probes GPT-5.6 file-deletion bug; Brex releases CrabTrap; Databricks valued at $188B.

Moonshot’s Kimi K3 nears 3T params and full weights by July 27; Capital One open-sources VulnHunter under Apache 2.0; Apple sues OpenAI, risking the company’s IPO timeline.

OpenAI sells a $230 Codex keyboard; GPT-5.6 Sol disproves a stat conjecture; Guardrails Alliance FEC filing appears July 15; Inkling open-weights live on Tinker; Gemma 4 CPU PR awaits review.

OpenAI's screenless speaker leak and lawsuit response, ACRouter's 2.6x cost win, and 1Password's AI Spend preview for fall 2026.

Waze adds Gemini chat and personalized routes; Tracebit shows context-bombing cuts agent compromises; Google publishes SensorFM wearable model; Anthropic localizes Claude pricing in India; Apple sues OpenAI.

Ploy migrates agents to GPT-5.6; Grok’s CLI uploads whole repos to GCS; Germany’s Soofi S open 30B tops open-model benchmarks; Claude Code adds a browser; DeepSeek cuts prices but token amplification persists.

Google's TabFM brings zero-shot tabular inference to BigQuery; OpenAI hires for family experiences; EU tells Meta to curb autoplay and infinite scroll.

Meta ships Muse Spark 1.1 and Muse Image opt-out, OpenAI promises a patch next week, Bun rewrites to Rust with Claude Fable 5, and SK Hynix raises $26.5B.

OpenAI is sued by Apple; ChatGPT Work with GPT-5.6 rolls out to paid tiers; Sol self-post-trained Luna; Hugging Face argues for open models; VB warns of an agent evaluation gap.

OpenAI ships GPT-Live full-duplex voice and ChatGPT Work on GPT-5.6; Meta begins MTIA chip production in September; Grok 4.5 ships; study finds 2.25x co-failure undercount.

OpenAI ships GPT-5.6 July 9 after Commerce approval; ZML launches free LLMD for multi‑chip inference; SambaNova raises $1B and gears SN50 shipping for H2 2026.

Meta ships Muse Image with opt-out tagging; Anthropic's J-lens reveals Claude's workspace; Microsoft moves Copilot to MAI; Cohere open-sources Arabic ASR.

Tencent ships Apache 2.0 Hy3 sized for export-compliant silicon; Zhipu launches ZCode; Vercel pushes model/agent split; Microsoft cuts ~4,800 roles; Expedia maps agent-release gates.

AI moves from models to practice: discovery demands in litigation, Mistral’s enterprise and open-weight plans, DiscoBench for search agents, and Anthropic’s drug R&D.

OpenAI's agent-first vision meets a 26,000-student study showing AI can cut exam performance by up to 24% over two years.

Today: Google???s June AI releases, Meta eyed compute resale, Gemini Spark on Mac, Meta agent setbacks, and Cloudflare crawler defaults affecting training data.