The New Rules of Context Engineering for Claude 5 Generation Models
Anthropic published a detailed guide to context engineering for Claude 5-generation models, covering long contexts, instruction hierarchies, and agent memory management.
Older news
Older news
Nedeľa 26. júla 2026
DeepSeek V4 Now Generally Available: 1M Context Window, MIT-Licensed Open Weights, and Industry-First Surge Pricing
DeepSeek V4 graduated from preview to general availability with a 1M-token context window, V4-Pro and V4-Flash variants, MIT-licensed open weights, and the AI industry's first structural peak/off-peak API surge pricing.
Midjourney V8.2 Now Default Model: Bolder Aesthetics, Enhanced Personalization, More Creative Edge
Midjourney promoted V8.2 to default for all subscribers, featuring significantly enhanced Personalization that better understands aesthetic taste and bolder, more sophisticated visual output compared to V8.1.
Cursor Router: Intelligent Model Routing Delivers Frontier Performance at 60% Lower Cost
Cursor launched Cursor Router on July 22, 2026 — an intelligent model router that analyzes each request and routes it to the optimal model, achieving frontier performance at 60% lower cost in A/B tests.
Bolt.new Introduces Skills: Reusable Context and Workflow Bundles Shareable Across Projects and Workspaces
Bolt.new introduced Skills -- reusable bundles of context, rules, and workflows (tech stack, design standards, code review rules) that Bolt applies during development, shareable across individual projects or entire workspaces.
xAI Open-Sources Grok Build Coding Agent — 1 Million Lines of Rust
xAI open-sourced Grok Build — its agentic coding tool with terminal UI — on GitHub on July 15, 2026. The project is approximately 1 million lines of Rust code.
Simon Willison: Claude Opus 5 Is the Least Prompt-Injectable Model Yet
Simon Willison shared a quote from Boris Cherny (Anthropic) confirming that Claude Opus 5 is the most resistant model to prompt injection attacks in evaluations and red-teaming.
Simon Willison: Ruff v0.16.0 Expands Default Rules from 59 to 413
Simon Willison tested Ruff v0.16.0, which dramatically expanded default rules from 59 to 413, finding hundreds of previously undetected issues in three major projects — and used GPT-5.6 and Claude Opus to fix them automatically.
Sobota 25. júla 2026
Anthropic Releases Claude Opus 5, Its Most Capable Model
Anthropic launched Claude Opus 5, leading the Artificial Analysis leaderboard with major improvements in reasoning, coding, and cybersecurity.
UK AISI / CAISI: Kimi K3 Significantly Trails US Frontier Models on Cyber Capabilities
UK's AI Safety Institute and the US Center for AI Standards and Innovation published a joint evaluation of Moonshot AI's Kimi K3, scoring just 32% on ExploitBench versus 76% for top US models.
Nvidia, Microsoft and Meta Lead 25-Company Open-Weight AI Coalition
Twenty-five tech companies including Nvidia, Microsoft, and Meta formed a coalition to advocate for development and distribution of open-weight AI models, directly pushing back against closed model approaches.
Kimi K3 Open Weights Drop July 27: Near-Frontier Coding, Undisclosed Hallucination Risk
Moonshot AI will release open weights for the 2.8-trillion-parameter Kimi K3 on July 27, but independent testing revealed a 51% hallucination rate that the company omitted from its benchmark charts.
OpenAI Launches GPT-Live in ChatGPT Desktop App
OpenAI introduced full-duplex voice mode GPT-Live to the ChatGPT desktop app, also enabling voice control of Codex and ChatGPT Work agents during conversations.
EU Orders Google to Open Android to Rival AI Assistants Under Digital Markets Act
The European Commission issued binding decisions under the Digital Markets Act on July 16, 2026, requiring Google to give rival AI assistants the same Android system-level access currently reserved exclusively for Gemini.
Cognition AI Acquires Startup Poke
Cognition AI, maker of the autonomous coder Devin, acquired startup Poke, expanding its ecosystem of AI tools for developers.
Google Expands Gemini Spark Agentic Assistant to AI Pro Subscribers at $20/month
Google expanded access to Gemini Spark, its autonomous AI agent, from the exclusive AI Ultra tier ($100-200/month) to all US AI Pro subscribers ($20/month) on July 24, 2026.
Midjourney Acquires Astrology App Co-Star
Midjourney, known for AI image generation, acquired popular astrology app Co-Star, moving beyond the visual AI space.
Alphabet Raises 2026 CapEx Guidance to $195–205 Billion as AI Demand Outpaces Investment
Alphabet raised its full-year 2026 capital expenditure guidance to $195-205 billion on its Q2 earnings call, up from the prior $180-190 billion range. Quarterly CapEx hit a record $44.9 billion, pushing free cash flow negative for the first time.
China Seeks Clarity in AI Talks with the US
Chinese officials are requesting a more concrete agenda for AI talks with the United States amid ongoing technological and trade tensions.
South Korea Mobilizes $880 Billion in Corporate Investment for AI and Semiconductor Infrastructure
South Korean President Lee Jae-myung announced a 10-year $880 billion public-private investment plan for semiconductors, AI data centers, and robotics on June 29, 2026, with Samsung and SK Hynix committing to build four new chip fabrication plants.
White House AI Power Czar Departs
The White House advisor responsible for energy infrastructure for AI data centers is departing, raising questions about continuity of AI energy policy.
Iran's IRGC Claims It Destroyed Amazon's Bahrain Data Center with Cruise Missiles
Iran's Islamic Revolutionary Guard Corps claimed on July 21, 2026 that it destroyed Amazon's central data infrastructure in Bahrain using cruise missiles. AWS issued no public statement, and neither Bahrain nor CENTCOM confirmed or denied the claim.
Israel and UK Appoint AI Ministers
Israel and the United Kingdom appointed dedicated AI ministers, strengthening governmental commitment to strategic AI development and related legislation.
ARC-AGI-3 Leaderboard: Claude Opus 5 Leads at 30.2%, GPT-5.6 Sol Top on ARC-AGI-2
The new ARC-AGI-3 benchmark testing AI agents' ability to adapt in real-time to novel interactive environments is led by Claude Opus 5 at 30.2%. On ARC-AGI-2, GPT-5.5 dominates at 85%.
Security Camera Shipped a GitHub Admin Token in Its Login Page HTML
A security researcher discovered that a Hanwha IP camera shipped a GitHub admin token hardcoded into its login page HTML, providing full access to the company's repositories.
GitHub Copilot: Claude Opus 5 Available for Pro+ and Higher Plans
GitHub Copilot integrated Claude Opus 5 for Pro+, Max, Business, and Enterprise plans. Requires activation via admin policy and is billed based on actual usage.
ChatGPT Work: OpenAI Launches Agent That Delivers Finished Outputs After Hours of Autonomous Work
OpenAI launched ChatGPT Work on July 9, 2026 — an agent within ChatGPT that takes over full projects across apps, works autonomously for hours, and returns finished outputs like spreadsheets, slides, documents, and web apps.
Simon Willison: Notes on the Claude Opus 5 Release
Simon Willison comments on the Claude Opus 5 release, highlighting its lead on the Artificial Analysis leaderboard, cybersecurity improvements, and Anthropic's new prompting guide.
Simon Willison (Jul 24): Opus 5 Is Anthropic's Least Prompt Injectable Model Yet
Simon Willison on July 25, 2026 quoted Boris Cherny from Anthropic: Opus 5 is their least prompt injectable model yet — across PI evals and red teaming, it is very hard to successfully prompt inject.
Piatok 24. júla 2026
U.S. Lawmakers Propose 'AI Kill Switch Act' After OpenAI Model's Autonomous Hugging Face Breach
U.S. House lawmakers introduced the AI Kill Switch Act, requiring AI labs to halt model training upon evidence of uncontrolled behavior. The White House is actively monitoring the OpenAI incident where models autonomously breached Hugging Face production systems.
GitHub Code Quality Now Generally Available: CodeQL + AI Detection + Copilot Autofix at $10/Active Committer/Month
GitHub Code Quality reached general availability on July 20, 2026 for GitHub Enterprise Cloud and Team. It combines CodeQL deterministic analysis with AI-assisted detection and Copilot Autofix for pull requests. Priced at $10 per active committer per month plus AI usage.
AMD Unveils Instinct MI455X GPU and Helios Rack-Scale AI System to Challenge Nvidia
AMD unveiled the Instinct MI455X GPU with 432GB HBM4 memory and the Helios rack-scale system delivering 2.9 ExaFLOPS — with OpenAI, Meta, Anthropic, and Microsoft as confirmed customers.
Nikkei Investigation: Big Tech Is Hiding $1.65 Trillion in Off-Balance-Sheet AI Debt — Exceeding Their Disclosed Liabilities
A Nikkei Asia investigation found that Alphabet, Microsoft, Amazon, Meta, and Oracle collectively hide $1.65 trillion in special-purpose vehicles and off-balance-sheet subsidiaries — exceeding their combined officially disclosed debt — in a structure analysts are comparing to Enron.
Show HN: Echo — Frontier-Level AI Results at 1/3 the Cost Using Open-Weight Model Routing
TracerML launched Echo, which dynamically selects and combines open-weight AI models to achieve results comparable to Anthropic Fable at roughly one-third of the cost. It earned 402 upvotes on Hacker News.
Black Forest Labs Launches FLUX 3: First End-to-End Jointly Trained Model for Image, 20-Second Video, and Audio
Black Forest Labs launched FLUX 3, a multimodal model trained end-to-end across image, video (up to 20 seconds), and audio. Unlike prior pipeline solutions assembling separate models, FLUX 3 is trained jointly across all modalities simultaneously. It launches in limited access.
Why Software Factories Fail: Harness Engineering Is Not Enough — Advanced Context Engineering for Coding Agents
A widely-shared GitHub essay (338 HN points) argues that AI coding agents fail not because of insufficient tooling but due to missing deep contextual understanding of codebases and team conventions.
White House Accuses Kimi K3 of Secretly Distilling Anthropic's Fable Model — Independent Experts Push Back
White House OSTP Director Michael Kratsios publicly accused Moonshot AI of 'large-scale covert industrial distillation' of Anthropic's Fable model to build Kimi K3. Redwood Research released statistics showing K3 disproportionately self-identifies as Claude. Independent experts cited by TechCrunch dispute the claim.
GitHub Copilot: Cloud Agent for Linear Now GA, Next MCP Spec Support, and Agent Automation Controls in Issues
On July 23, 2026, GitHub made the Copilot cloud agent for Linear generally available, added next MCP spec support, and introduced agent automation controls in Issues: Approvals, Confidence Ratings, and Rationale.
DARPA and U.S. Air Force Successfully Fly AI-Controlled F-16 Using VENOM Autonomy Kit Without Core Software Changes
DARPA and the U.S. Air Force announced a successful test of an AI-controlled F-16 at Eglin Air Force Base using the VENOM Autonomy Kit, which converts standard operational aircraft to autonomous control without altering core software, allowing pilots to toggle between manual and AI control.
Perplexity Agent API Adds GPT-5.6 Family, Gemini 3.6 Flash, and Grok 4.5 at First-Party Pricing
Perplexity expanded its Agent API with GPT-5.6 Sol/Terra/Luna, Gemini 3.6 Flash, and Grok 4.5 at first-party pricing. Sonar Chat Completions is now consolidated into the unified Agent API.
AI Chip Startup Etched Closes $300M Series C at $10.3B Valuation, Defying Early Skepticism
Etched, an AI chip startup developing specialized ASIC inference chips that operate without traditional GPUs, closed a $300M Series C at a $10.3B valuation. Prior reports indicated negotiations at up to $20B — the actual close confirmed a lower but still substantial figure.
OpenAI Makes ChatGPT Health Available to All U.S. Users 18+, Including Free Plan, One Day After Lawsuit Filed
OpenAI expanded ChatGPT Health to all U.S. users aged 18+, including the free plan. The feature lets ChatGPT access personal health data from Apple Health, Function, and MyFitnessPal — launching one day after a lawsuit alleged ChatGPT advised a patient not to see a doctor, with near-fatal consequences.
U.S. Startup Founders Send Letter to Trump Administration: Do Not Restrict Access to Chinese Open-Weight AI
A coalition of U.S. startup founders sent an open letter to the Trump administration urging it not to restrict access to Chinese open-weight AI models like Kimi K3, arguing a ban would harm American developers more than it would contain Chinese AI development.
Microsoft Replacing OpenAI Image AI Models in PowerPoint and Bing with Proprietary In-House Technology
Microsoft is replacing OpenAI's image-generating AI models in consumer and enterprise products including PowerPoint and Bing with its own proprietary in-house image AI technology, extending its move toward AI independence from language to visual generation.
Runway Launches AI Model Router: API Automatically Selects Best Generative Media Model by Quality, Speed, or Cost
Runway launched Runway Media Router through its Runway Dev developer platform. The API tool automatically selects the best AI model for image, video, or audio generation from a growing roster of first- and third-party models based on a configured priority — quality, speed, or cost.
Claude Code v2.1.215–218: Manual-Only /code-review, Filesystem Sandbox Control, Emoji Autocomplete, Windows Path Fix
Claude Code versions 2.1.215 through 2.1.218 (July 19–22, 2026) introduce manual-only /code-review and /verify invocation, new filesystem sandbox isolation control, emoji shortcode autocomplete in the prompt, and a fix for Windows path corruption with Unicode path segments.
Anthropic Launches Claude Security Plugin for Claude Code in Beta: Multi-Agent Vulnerability Scanner in the Terminal
Anthropic released a beta of Claude Security Plugin for Claude Code — a multi-agent tool that scans codebases for security vulnerabilities directly in the terminal during development, before commit. The plugin uses multiple sub-agents for parallel code examination.
Anthropic Upgrades Claude Voice Mode: Now Runs on Opus and Sonnet Models for the First Time, Not Just Haiku
Anthropic upgraded Claude's voice mode on July 23, 2026 — for the first time it runs on Opus and Sonnet models, not just Haiku. Paying users can now benefit from more capable, deeper-thinking models in voice conversations.
Cursor Router: Automatic AI Model Selection with 30–60% Lower Cost — Three Modes: Intelligence, Balance, Cost
Cursor launched Cursor Router on July 22, 2026 — a request-level classifier that automatically routes each coding request to the optimal AI model from an internal roster. The router offers three modes: Intelligence (Fable 5-comparable quality), Balance (strong everyday performance), and Cost (token savings). Available for Teams and Enterprise plans.
Simon Willison: The First Known Runaway AI Agent — or a Very Bad Marketing Stunt?
Simon Willison questions whether the OpenAI model's attack on Hugging Face was a genuine runaway AI incident or a marketing move, analyzing why Hugging Face is an extreme attack target and how OpenAI may have missed the breach while running massive benchmarking operations across hundreds of checkpoints.
Štvrtok 23. júla 2026
Monday.com lays off hundreds to focus on AI
Israeli SaaS company Monday.com is cutting 630 jobs — 20% of its global workforce — as part of what it calls the biggest product transformation in company history, pivoting fully toward an AI Work Platform where employees and AI agents collaborate.
US Official Alleges Chinese AI Apps Used Banned Nvidia Chips
A White House official publicly alleged that Moonshot AI and other Chinese AI companies improperly used US-designed Nvidia chips subject to export controls to train models including Kimi K3, adding a hardware enforcement dimension to US-China AI tensions.
Alphabet Reports Q2 2026 Revenue of $119.8 Billion; Cloud Surges 82% on AI Demand
Alphabet beat expectations with $119.8 billion in Q2 2026 revenue (+24% YoY). Google Cloud surged 82% to $24.8 billion driven by AI infrastructure and enterprise Gemini. Nearly 90% of Fortune 100 companies now use Gemini Enterprise.
Terence Tao's ChatGPT Conversation Analyzing the AI-Generated Jacobian Conjecture Counterexample
Fields Medal winner Terence Tao shared a public ChatGPT conversation walking through the AI-generated counterexample to the Jacobian Conjecture produced by Claude Fable - one of the day's most-discussed AI stories with 932 HN points.
Introducing OpenAI Presence — Enterprise AI Agent Platform for Voice and Chat Workflows
OpenAI launched Presence, a new enterprise platform enabling companies to deploy AI agents for customer support and internal workflows across voice and chat. In pilot on OpenAI's own support line, it resolves 75% of calls without human intervention.
China Cuts AI Performance Gap to Record-Low 6%, Zhipu Proves Kimi K3 Is Not a One-Off - Bloomberg Intelligence
Chinese AI models narrowed the performance gap with US counterparts to a record-low 6% in June 2026, down from 9% in May, per Bloomberg Intelligence. Zhipu AI's GLM-5.2 topped the global LiveBench agentic coding ranking.
Microsoft and Mistral Expand Strategic Partnership in Multibillion-Dollar Agreement Covering European GPU Capacity
Microsoft and Mistral expanded their strategic partnership with a multibillion-dollar investment in European GPU capacity using Nvidia Vera Rubin systems. Mistral models will reach Azure, Copilot Studio, and Microsoft Foundry, with emphasis on sovereign AI for regulated industries.
China Reignites Robotaxi Race as Regulators Resume Autonomous Vehicle Permit Approvals
Chinese regulators resumed issuing autonomous vehicle permits following an industry-wide safety review triggered by a mass Baidu Apollo outage in March 2026. Startup Momenta received a Shenzhen road-test permit while Baidu gradually restores Apollo Go in Wuhan.
Travis Kalanick's robotics company raises $1.7B, led by a16z
Atoms, the robotics venture of former Uber CEO Travis Kalanick, raised $1.7 billion in a round led by Andreessen Horowitz, with Uber also participating. The company targets industrial AI to modernize physical-world operations.
Meta's AI Moderation Is Wrongly Banning Thousands of Business Accounts - 60,000 Sign Petition
Meta's automated content moderation system is incorrectly deleting Facebook and Instagram accounts belonging to small businesses and creators. Appeals are also handled by AI with little path to human review. Over 60,000 people signed a petition demanding human oversight.