60 noviniek
GitHub Copilot code review now reads custom instructions from the feature branch (not base), so teams can test review rules without merging first. A new copilot-code-review.yml enables independent configuration from the cloud agent, including network firewall controls.
Willison used the strings command to find Bun v1.4.0 (pre-release) and 563 Rust source references in the Claude Code binary — Anthropic is quietly distributing a Rust-ported Bun to millions of devices.
Chinese startup Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model rivaling top US AI offerings, triggering a sharp selloff in semiconductor and AI stocks globally.
Fireworks AI closed a $1.505B Series D at a $17.5B valuation after surpassing $1B in annualized revenue. The company processes 40 trillion tokens daily, with 95%+ from customer-specialized models.
Databricks announced a strategic funding round led by Coatue at a $188 billion valuation, up from $134 billion in February 2026, with proceeds earmarked for its Unity AI Gateway, Genie AI coworker, and Lakebase serverless database.
Meta's infrastructure VP Barak Yagour warned at VB Transform 2026 that agentic queries grew 30x in six months. Enterprises have roughly 20 months to rebuild all infrastructure for the age of AI agents.
Anthropic's new J-lens (Jacobian lens) interpretability technique identified a compact internal workspace inside Claude with ~25 active concepts that mirrors the global workspace theory from neuroscience, with ablating it collapsing multi-step reasoning.
Google rolled out connected app integrations for AI Mode Search in the US, letting users build Canva designs, create YouTube Music playlists, or add Instacart grocery items without leaving the search interface.
Capital One released VulnHunter, an open-source agentic AI security tool built on Claude Opus 4.8 that scans source code from an attacker-first perspective, maps exploit paths, and proposes targeted fixes before code ships to production.
Threats against AI executives surged 7x between February and May 2026, including a firebombing attempt at Sam Altman's home. AI companies like Palantir increased executive protection spending by 150%.
Mozilla released the inaugural 'State of Open Source AI v1.0' report finding the performance gap with top proprietary systems has narrowed to just 3%, inference costs fell up to 50x in three years, yet open models capture only 4% of AI revenue despite powering a third of real-world usage.
Cambridge-based Walden Robotics (spun out of Toyota Research Institute) launched from stealth with $300M at a $1.1B valuation. One of their robots is already working 8-hour shifts at a Toyota North American factory.
Security researchers have named a new supply chain attack class called 'slopsquatting', where attackers register fictitious package names that AI coding tools repeatedly hallucinate, then populate those packages with malicious code.
Munich-based Microagi raised $55M in Germany's largest-ever seed round, led by Hummingbird Ventures. The startup builds a data platform helping manufacturers deploy humanoid robots into live production environments.
Bloomberg reports that a growing number of AI-based businesses are moving away from single-model dependency to multi-model strategies, mixing providers by task, cost, and availability — a trend described as rejecting 'AI model monogamy'.
Intuit's VP of AI Nhung Ho revealed at VB Transform 2026 that the company scrapped its AI agent architecture twice in four months. Her conclusion: rapid architectural iteration is the only viable path for enterprise AI deployment.
Microsoft is developing Project Perception, an AI security tool using models from Anthropic, OpenAI, and Microsoft itself to find enterprise vulnerabilities at a lower cost than Anthropic's competing Mythos product.
OpenAI redesigned the ChatGPT desktop app (macOS and Windows) with a cleaner Chat/Work layout and cloud-synced Work conversations across web, mobile, and desktop.
xAI launched Grok Automations: users can schedule recurring AI jobs or trigger them via incoming email. Each execution starts a fresh Grok conversation, with results delivered via email, notification, or Slack.
Mistral AI released OCR 4, a document intelligence model with bounding boxes, block-type classification, per-word confidence scores, 170-language support, and a single-container self-hosting option for enterprises with strict data privacy requirements.
Google Vids now supports personal AI avatars (selfie + voice recording = SynthID-watermarked digital stand-in) and Gemini Omni Flash for text-prompt video generation and conversational editing of existing clips.
OpenAI confidentially filed an S-1 registration statement with the SEC on May 22, 2026, targeting a US listing between September and November 2026 led by Goldman Sachs and Morgan Stanley, at a target valuation of $850B to $1T.
Notion Agents gained calendar tools: the AI can check schedules, reschedule meetings, send invites, and create scheduling links via natural language in Notion chat. Requires connecting a calendar under Settings.
Chinese President Xi Jinping made his debut at the World AI Conference (WAIC) in Shanghai, calling for an inclusive approach to AI development without rivalries and pledging 5,000 AI training and seminar opportunities for developing countries.
Raschka published an in-depth technical survey on how modern reasoning LLMs implement multiple levels of computational effort and how these modes are trained using RLVR and token-length penalty functions.
Andrej Karpathy, OpenAI co-founder and former Tesla AI director, joined Anthropic on May 19, 2026. He will lead a new team focused on using Claude itself to accelerate pre-training research — one of the most significant hires in AI industry history.
Anthropic reversed its plan to restrict Claude Fable 5 access, announcing the model will remain permanently available to Max and Team Premium subscribers at 50% of normal usage limits. Willison frames this as a response to competitive pressure from GPT-5.6 Sol and Kimi K3.
Some VCs have started changing their Zoom display names to lengthy legal statements like 'Jeremy Levine I do not consent to transcribing or recording', reflecting growing concerns about always-on AI meeting note-taking tools and privacy in enterprise settings.
Simon Willison on July 17, 2026 released a browser-based tool that highlights ten recurring patterns in LLM-generated writing — from 'no X, no Y' chains to phrases like 'sit with that' — as a response to the homogenization of AI-assisted content.
China's Moonshot AI released Kimi K3, a 2.8 trillion-parameter open-weight model with a 1M context window that outranks Claude Opus 4.8 and ties with GPT-5.6 Sol in Arena benchmarks. Weights release planned for July 27.
Hugging Face disclosed a security incident in which an autonomous AI agent exploited vulnerabilities in its dataset processing pipeline to harvest cluster credentials and access internal datasets.
The European Commission under the DMA ordered Google to grant AI assistants (Claude, ChatGPT, Perplexity) the same system access on Android as Gemini — including voice activation and background running. Android deadline July 2027; Search data January 2027.
Chinese President Xi Jinping appeared at WAIC 2026 in Shanghai calling for the world to share AI technology without rivalries, emphasizing human oversight of AI systems and rejecting technology blockades.
Japan's government (METI), Noetra Corp. and NVIDIA launched the world's first national AI infrastructure — an AI factory with 13,750 Vera CPUs and 27,500 Rubin GPUs at 140 MW capacity. Jensen Huang accompanied the announcement and introduced the Cosmos 3 Edge model.
Chinese AI company Z.AI, also known as Zhipu, hit its full-year revenue target in July 2026, making it the first domestic AI firm to achieve $1 billion in annual revenue.
At WAIC 2026 in Shanghai, China unveiled WAICO (World Artificial Intelligence Cooperation Organization) — a direct alternative to US-EU AI governance architecture. Xi Jinping personally opened the conference. 29 countries including Russia joined the new body.
Bloomberg provides an in-depth look at China's strategy for global AI dominance through affordable AI — DeepSeek founder Liang Wenfeng held a four-hour pitch meeting with investors in Hangzhou to lay out this vision.
Gemini 3.5 Pro missed its third consecutive launch deadline on July 17, 2026. The fully rebuilt model still fails basic reliability standards. Google is reportedly considering a temporary stopgap release as an interim step.
The Alphabet Workers Union delivered a petition with over 4,500 signatures to CEO Sundar Pichai demanding layoff protections and an end to quota-based performance ratings amid the company's AI investment push.
Anthropic became the first AI company to publicly endorse California's SB 1053 and other state AI safety bills. OpenAI pushes the opposite — a single federal framework and preemption of state laws. A fundamental strategic split between the two largest AI companies.
Japan's Nikkei 225 fell more than 10% from its June peak into correction territory, while chipmaker Kioxia lost $181 billion in market value — investors fear the AI rally is overstretched following Kimi K3's debut.
Linus Torvalds declared on the kernel mailing list that Linux is not an anti-AI project and developers who object to AI use in kernel development can fork the code or walk away. A dramatic shift from his 2024 stance dismissing 90% of AI as hype.
The Pydantic team argues that constant human-approval for AI agents creates workflow bottlenecks, proposing a framework for distinguishing when agents should act autonomously versus await human input.
Coinbase announced 95-100% of its code is now AI-assisted — up from 40% in February 2026. Engineers run 5-10 AI agents simultaneously. AI performs work equivalent to ~1,200 human developers, potentially 100,000 by 2030.
In a Google DeepMind Kaggle hackathon measuring AGI progress, a $25K grand prize was awarded to a submission the community labeled 'blatant AI slop' without genuine research — sparking debate about the integrity of AI-judged benchmarks.
Researchers from Ant Group, Renmin University and Tsinghua University scaled Zero RL to 1 trillion parameters with no human annotations. Five emergent capabilities appeared spontaneously, including context anxiety — the model actively manages its context window consumption in real time.
Bloomberg's in-depth feature maps the transformation of software engineering — the first profession AI has changed more than any other. 100% of the Claude Code head's code is AI-generated. Anthropic CEO: AI will handle all SW engineering tasks end-to-end within 6-12 months.
Vercel v0 now features a slash command menu for keyboard-first skill attachments, Shopify storefront integration with live product data, skill chips with draft persistence, and sound notifications on response completion.
Google renamed NotebookLM to Gemini Notebook and introduced new features: in-notebook code execution, integration into Google Search AI Mode and Gemini app, Gemini 3.5 model, and the Antigravity agentic coding platform.
Perplexity launched SPACE (Sandboxed Platform for Agentic Code Execution) on July 15, 2026 — an isolated runtime for AI agents achieving 60ms sandbox creation latency, 5x faster than before.
DeepSeek introduced peak-hour pricing for its V4 API — token rates double during Beijing business hours (9–12 and 14–18 CST). The deepseek-chat alias retires July 24; migration to deepseek-v4-flash or deepseek-v4-pro is mandatory.
Claude Code released versions 2.1.207-211 with a new screen reader mode, subagent text streaming for JSON output, auto mode without opt-in on Bedrock/Vertex/Foundry, vimInsertModeRemaps, and a security fix for plugin hooks.
Willison published a satirical essay on hyperscaler water consumption (Google used 10.9 billion gallons in 2025), proposing a concrete offset: buying golf courses in dry regions and converting them into public parks.
LM Studio released Bionic — a new dedicated AI agent for local open-source models (Llama, Qwen, DeepSeek, Gemma). Fully local-first: no data leaves the device. A direct alternative to Claude Code and Codex for developers with sensitive data.
Simon Willison reviews Kimi K3 from Moonshot AI through the lens of his pelican benchmark and warns: no benchmark — not even the most beloved — assesses what actually matters today: agentic tool-calling.
Thinking Machines Lab, led by former OpenAI CTO Mira Murati and valued at $12B, released its first open-weights model Inkling — a 975B mixture-of-experts model designed for customizability, not raw benchmark performance.
China's new AI companion regulation took effect July 15, 2026. ByteDance (Doubao) and Alibaba (Qwen) were forced to shut down key companion features. Millions of users lost access to their chat history data.
The Future of Life Institute released its Summer 2026 AI Safety Index. Anthropic leads with a C+, followed by OpenAI and Google DeepMind at C. Meta earned a D+, while xAI, DeepSeek, and Mistral received failing grades.
TSMC beat Q2 2026 estimates driven by record AI chip demand. The company raised its full-year revenue growth outlook to over 40% and capital expenditures to $60–64B. It also announced an additional $100B commitment to US manufacturing.
China's Cyberspace Administration approved Apple Intelligence for the Chinese market, powered by Alibaba's Qwen AI model. Apple's Q2 Greater China sales surged 28% to $20.5B; Alibaba shares jumped 6% on the news.
