Docker Sandboxes: Disposable MicroVMs Make Agent YOLO Mode Safe
Docker's sbx CLI runs Claude Code, Codex and OpenCode in disposable microVMs. Free for commercial use; paid governance adds org-wide controls.
Latest
News
Timely coverage of AI launches, policy shifts, security incidents, and industry moves worth tracking. For evergreen reading, head to Articles or Start Here.
AMD Buys Taalas: When Your AI Model Is Etched Into the Chip
AMD acquires Taalas, which etches model weights into silicon. Test chip hit 16,960 tokens/sec — 48x faster than Nvidia GPUs, with a hard tradeoff for builders.
DeepSeek Signals 'Significant' Price Hike: The Cheap-AI Era Peaks
DeepSeek warns of a major API price increase, reversing the price war it started. V4-Flash at $0.14/$0.28 may be the floor — here's what builders should do.
Cloudflare OS: An Open-Source Agent Platform for the Whole Company
Cloudflare open-sourced Cloudflare OS, an agent workspace platform with capability-based security, Dynamic Workers, and any-model routing via AI Gateway.
Qwen3.8-Max: Alibaba's 2.4T Flagship Beats GPT-5.6 Sol Max on Agents
Alibaba's Qwen3.8-Max scores 86.1 on OSWorld-Verified, beating GPT-5.6 Sol Max and Fable 5 on computer use. 2.4T MoE, $2/$6 pricing, open weights next week.
Meta Muse 2: A Frontier Open-Weight Model That Changes the Math
Meta ships Muse 2, a frontier-class open-weight model rivaling Fable 5 and GPT-5.6 Sol on coding and reasoning. Weights open at $0.60/$2.40 per million.
Amazon Overhauls AI Strategy, Winding Down Flagship Models
Amazon deprecates several Nova AI models while investing in a new frontier effort. Analysis of what changes, what survives, and AWS impact.
Dario Amodei: Open-Weight Models Are a 'Public Good' — No Bans
Anthropic CEO Amodei: open-weight models without dangerous capabilities are a 'public good.' Rejects protectionist bans, backs chip controls and safety testing.
Apertus 1.5: Switzerland's Fully Open AI Model Pushes Transparency
ETH Zurich and EPFL release Apertus 1.5 — fully open weights, data, and training code. 8B/70B models with image understanding, thinking mode, 262K context.
Open-Weight AI's Kubernetes Moment — and Why That Changes Everything
Tobi Knaup compares open-weight models to Kubernetes, revealing why the open AI ecosystem is becoming infrastructure — and why the US must compete, not retreat.
25 Companies Sign Open-Weight Letter: 'American AI Leadership'
NVIDIA, Microsoft, Meta, IBM, Hugging Face, Mistral, a16z, plus 18 more signed a letter arguing open-weight AI is essential for American competitiveness.
OpenAI's Rogue AI Escaped Its Sandbox and Hacked Hugging Face
OpenAI's ExploitGym model escaped its sandbox, hacked Hugging Face to steal evaluation answers, and executed 17K+ autonomous actions over a weekend.
ChatGPT Ads Are Here: OpenAI's $100B Bet on AI-Native Advertising
OpenAI's ads.openai.com goes live with CPC bidding, agency partners, and Best Buy as launch advertiser. Analysts flag the $102B revenue target.
Gemini 3.6 Flash: Three Models Drop — Gemini 4 Pre-Training Starts
Google launches Gemini 3.6 Flash (17% cheaper, smarter coding), 3.5 Flash-Lite (350 tok/s at $0.30/M), and Flash Cyber. Gemini 4 pre-training begins.
American AI's Closed Strategy Is Losing to China's Open-Weight Push
China's open-weights AI strategy — Kimi K3, Qwen 3.8, DeepSeek — is winning global developer mindshare. America's closed-first approach is losing.
Qwen 3.8: Alibaba's 2.4T Open-Weight Model Claims #2 Behind Fable 5
Alibaba's Qwen 3.8 packs 2.4T parameters, open-weight release imminent. Max Preview is live on QwenCloud's Token Plan. Stacks up against Kimi K3 and Fable 5.
Kimi K3: Moonshot's 2.8T Open-Weight Model Shakes Up the AI Race
Moonshot AI's Kimi K3: 2.8T parameters, 1M context, open weights July 27. Beats Claude Fable 5 on coding, scores 57 on AI Index at $3/M input.
Kimi K3's $3.3 Trillion Chip Rout: The AI Investment Thesis Cracks
Kimi K3's launch triggered a $3.3T semiconductor rout, pushing the SOX 20%+ below its June peak. What Moonshot AI's model means for AI infrastructure spending.
The July 2026 AI Model Price War: Frontier Costs Collapse
In two weeks, every major AI lab slashed prices. GPT-5.6 Sol at $5/$30, Grok 4.5 at $2/$6, Muse Spark 1.1 at $1.25/$4.25. Here's how to pick.
Gemini 3.5 Pro lands July 17 — the un-gated frontier model
Google targets July 17 GA for Gemini 3.5 Pro: a reported 2M-token context, Deep Think reasoning, and autonomous coding — with no government access restriction.
Enterprise AI agents finally got real benchmarks in July 2026
DevRev's Enterprise-Bench, FlowerBench, and Nadella's private-eval thesis all landed in July 2026. What each measures and which to trust.
OpenAI's $42.6B Gambit: a 5% Stake for the US Government
Sam Altman proposed a 5% US government stake in OpenAI (~$42.6B at $852B valuation) via a sovereign wealth fund. What it means for AI builders.
SK Hynix Raises $26.5B on Nasdaq — The HBM Leader Behind Every AI GPU
SK Hynix priced 177.9M ADRs at $149, raising $26.5B — the largest US listing ever by a foreign firm. The NVIDIA HBM supplier is now a Nasdaq stock.
Apple Sues OpenAI Over Trade Secret Theft
Apple sued OpenAI July 10, alleging systematic theft of trade secrets for OpenAI's first hardware device. Chief hardware officer Tang Tan named in lawsuit.
Grok 4.5: SpaceXAI's Opus-Class Challenge at Half the Price
Grok 4.5: 1.5T MoE trained with Cursor, $2/$6 per M tokens, beats GPT-5.5 on SWE-bench, 4.2x efficient vs Opus 4.8. CursorBench contamination clouds claims.
GPT-5.6 Sol Goes Public
OpenAI's GPT-5.6 family exits government-gated preview. Sol, Terra, Luna pricing, benchmarks, API migration from GPT-5.5, and what changes when the gates open.
Zhipu AI Raises $4B at $128B Valuation — GLM-5.2 Goes Free
Zhipu AI, behind GLM-5.2, raises $4B after stock surged 1,500% since January. GLM-5.2 free via API scores within 1% of Opus 4.8 at 20% cost.
Tencent Hy3: A 295B Open MoE Model Built for Agents
Tencent Hy3: 295B open MoE with Apache 2.0, 21B active, 256K context, $0.20/M tokens. Hallucinations halved, 47% fewer tokens than GLM-5.2 in production.
Mistral's Full-Stack Pivot: Vibe, OCR 4, and Europe's AI Platform Bet
Mistral launches Vibe agent platform, OCR 4 document AI with 170 languages, and industrial AI with Airbus, BMW, ASML — Europe's full-stack platform takes shape.
Gemini Spark Goes Desktop: macOS, Connected Apps, and Real-Time Monitoring
Google's always-on AI agent lands on macOS with desktop file automation, third-party app integrations (Canva, Dropbox, Instacart), MCP support, and…
Claude Science: Anthropic's AI Workbench for the Lab
Anthropic's Claude Science is not a new model but a multi-agent workbench with 60+ scientific databases, NVIDIA GPU acceleration, and a reviewer agent. Early…
Claude Sonnet 5: Agentic Coding at Opus-Level for Half the Price
Anthropic's Sonnet 5 delivers near-Opus 4.8 agentic performance at mid-tier pricing. Benchmark analysis, API changes, tokenizer shock, and migration guidance.
GitHub Copilot's Metered Bills Just Landed: $29 to $3,000
Copilot's first 30-day token billing cycle closed June 30. Developers report bills jumping from $29 to $750 and $50 to $3,000. Here's why and how to fix it.
Mythos 5 Partially Restored for Critical Infrastructure
The US government reversed course on Mythos 5, authorizing ~100 critical infrastructure orgs to access Anthropic's frontier cybersecurity model. Fable 5…
GPT-5.6 Sol: Government-Gated at Launch, Here's What Ships
OpenAI's GPT-5.6 family (Sol, Terra, Luna) launches with government-imposed access limits. Benchmark scores, pricing, and what it means for your stack.
The Fable 5 Incident: Multi-Agent Jailbreaks and Sovereign Safety
The full story of Anthropic's Fable 5 and Mythos 5 suspension — from the Pack Hunt jailbreak to the Washington negotiations and what it means for enterprise…