US/UK labs · flagship models

FRONTIER

Chinese labs · models · geopolitics

CHINA AI

the biggest threads

CODING & AGENTS

ecosystem · regulation · risk

OPEN & SAFETY

money · deals · execs

BUSINESS

AI in verticals

INDUSTRIES

chips · infra · power · local

HARDWARE

image · video · audio

MEDIA & VOICE

American artificial intelligence corporation

American artificial intelligence research organization

American multinational technology company, a subsidiary of Alphabet Inc.

country located primarily in North America

AI language model and chatbot developed by Anthropic for API access

American multinational technology company

family of language models by Google DeepMind

American multinational technology corporation

Anthropic's frontier model in a new 'Capybara' tier sitting above Opus, first revealed via a leaked draft blog post in March 2026.

Anthropic's flagship Claude model tier, used for high-capability reasoning and coding tasks.

Open-source personal AI agent created by Austrian developer Peter Steinberger; acqui-hired by OpenAI in February 2026 to bolster ChatGPT's personal-agent surface.

American internet infrastructure and website security company

artificial intelligence company owned by Google

AI-native code editor and the company behind it (Anysphere).

American multinational technology company based in Cupertino, California

OpenAI's coding-agent product family. GPT-5.1-Codex-Max (November 2025) is the long-horizon agentic-coding tier; Codex-Mini is the cost-efficient tier.

artifical intelligence model geared towards programming

CEO of OpenAI; co-founder of Y Combinator alumni network and key public face of the AI frontier-lab era.

Google's family of open-weight AI language models with multi-token prediction support for faster inference

artificial intelligence chatbot developed by OpenAI

Amazon Web Services; cloud infrastructure subsidiary, builder of Trainium AI chips and the Nova foundation-model family.

family of large language models by Google

artificial intelligence chatbot developed by Alibaba Cloud

Chinese multinational artificial intelligence company

Anthropic's internal vulnerability-discovery programme that tests Claude Mythos against real codebases.

DeepSeek's upcoming ~1 trillion parameter mixture-of-experts model, designed to run on Huawei Ascend chips.

xAI's family of large language models accessed primarily through X (Twitter) and the Grok API.

Open standard from Anthropic (Nov 2024) that lets AI assistants connect to external tools and data sources via a uniform interface; first endorsed by a major platform vendor (Microsoft Copilot Studio) in May 2025.

Amazon.com, Inc.; parent of AWS and the Nova foundation-model family.

headquarters of the United States Department of Defense; named after the shape of the building, a polygon with five (greek: penta) sides

Mid-tier model in Anthropic's Claude family, balancing capability and cost for everyday production workloads.

OpenAI's flagship model launched August 7 2025; unified router architecture with fast and reasoning paths, 400K context, ~half GPT-4o input cost.

Moonshot AI's Kimi family of large language models, including Kimi K2 and K2.5.

political and economic union of 27 European states

AI research organization developing open-source language models and agent systems

cloud-based services and infrastructure from Google

Cloudflare Workers feature enabling stateful, persistent execution for serverless applications

protocol for applications to provide context to LLMs

Cloudflare software development kit for building AI agents, including the McpAgent class for remote MCP servers

Google's AI-powered research and notebook product that summarises and chats with user-uploaded sources.

Large language model developed by DeepSeek, used as a math benchmark comparison point

country in Western Europe and other continents (through its overseas territories in America, Africa and Oceania)

Meta's frontier AI research division led by Chief AI Officer Alexandr Wang, responsible for the Muse Spark model.

UK government institute evaluating AI safety and security risks, part of DSIT

Anthropic's small/fast Claude tier; Claude Haiku 4.5 (Oct 2025) added hybrid reasoning at / per million tokens.

Japanese multinational holding conglomerate; one of OpenAI's largest investors and a Stargate Project anchor partner.

Nous Research's family of open-weight self-improving agentic LLMs.

Chinese multinational investment holding conglomerate

Anthropic's AI assistant and large language model product

Cloudflare product providing AI inference capabilities on the Workers edge runtime platform

American conversational search engine company

AI research lab founded February 2025 by ex-OpenAI CTO Mira Murati; raised B at B in July 2025 led by Andreessen Horowitz; shipped Tinker fine-tuning API in October 2025.

Microsoft-owned developer platform and code-host; launched Agent HQ at GitHub Universe 2025 to orchestrate coding agents across multiple AI labs.

GitHub's AI coding-assistant product; expanded in October 2025 (Universe) to host third-party agents from Anthropic, OpenAI, Google, Cognition, xAI under a single subscription.

private research university in Stanford, California, USA

American multinational mass media company

software for running large language models (LLMs) on a local computer instead of in cloud servers

British artificial intelligence researcher (born 1976)

Internal codename for OpenAI's next frontier model after GPT-5, expected to ship as GPT-5.5.

Anthropic's suite of composable APIs for building and deploying cloud-hosted Claude agents at scale.

Anthropic's prompt-to-design surface for slide decks, app prototypes and marketing pages, powered by Claude Opus.

American legal AI startup building generative AI tools for law firms and in-house legal teams.

Anthropic's collaborative AI tool for team-based work integrated into Claude Desktop

Cloudflare product providing a unified API gateway for AI services and model providers

source code editor developed by Anysphere

Cloudflare managed retrieval-augmented generation primitive for building AI agents on Workers platform

Irish-American payment technology company

Google DeepMind's generative video model family; Veo 3 launched at I/O 2025 with higher fidelity and improved prompt adherence.

artificial intelligence platform developed by Apple

American multinational AI infrastructure investment joint venture

CEO of Microsoft AI; co-founder of DeepMind, ex-CEO of Inflection AI; leading Microsoft's in-house frontier-model effort (MAI-1, MAI-Voice-1).

Founder and CEO of Thinking Machines Lab; former CTO of OpenAI.

US AI lab founded March 2024 by ex-DeepMind researchers Misha Laskin and Ioannis Antonoglou; raised B at B in October 2025 led by Nvidia to build a Western open-weight frontier alternative.

private university in Cambridge, Massachusetts, USA, founded 1861

American multinational software/cloud company; key Stargate Project partner with OpenAI on AI data-centre buildouts.

JavaScript runtime created by Jarred Sumner; bundled in Claude Code from July 2025 and acquired by Anthropic in December 2025.

🤖 Cloudflare ships agents-sdk — a JavaScript framework for building stateful AI agents on Workers

🔌 Cloudflare opens remote MCP servers — with an OAuth library so they don't have to live on your laptop

🆓 Cloudflare puts Durable Objects on the free plan — and ships AutoRAG and Workflows GA in the same drop

🪜 Cloudflare ships 13 first-party MCP servers — plus a 10-partner Demo Day (Atlassian, Linear, Stripe, Anthropic)

🪨 Anthropic ships Claude 4 — Opus 4 and Sonnet 4 lead with coding and long-horizon agents

🤖 Google I/O reframes Gemini around 'agentic AI' — agent mode, Project Astra in Live, AI Mode in Search

🎬 Veo 3, Imagen 4, and Flow — Google's new media-gen wave

🛠️ Jules — Google's autonomous coding agent enters public beta

🏗️ Microsoft Build — Copilot Tuning, multi-agent orchestration, and 11K models in Foundry

🔌 Microsoft endorses MCP — Anthropic's protocol gets first major platform adoption

🧮 AlphaEvolve breaks Strassen's 1969 record — DeepMind's evolutionary coding agent finds better algorithms

📱 NotebookLM gets a mobile app — offline Audio Overviews and on-the-go source loading

💸 Moonshot AI raises ~$2B at a $20B valuation — China's open-source bet pays again

🐳 DeepSeek slips R2, ships R1-0528 update instead

🥺 OpenAI rolls back GPT-4o sycophancy update — first public model-behaviour postmortem

🍎 Apple WWDC — Foundation Models framework opens on-device LLMs to developers

🧠 OpenAI ships o3-pro — slower, deeper, beats Claude 4 Opus on PhD-level science

⚡ Mistral ships Magistral — Europe's first open-weight reasoning model

💼 Meta pays $14.3B for 49% of Scale AI — Alexandr Wang becomes Meta's first Chief AI Officer

💸 Cursor closes $900M Series C at $9.9B valuation — coding-agent ARR going vertical

⚖️ Bartz v. Anthropic — Judge Alsup splits AI training fair use from piracy

📞 Apple weighs replacing Siri's brain with Anthropic or OpenAI

🧪 The Browser Company ships Dia — an agentic browser around your tabs

📱 OpenAI/Jony Ive AI device leaks — screenless, pocket-sized, target 2026

🧬 Zuckerberg pulls a Manhattan Project — Meta's $100M+ talent raid for Superintelligence Labs

⚖️ NYT v. OpenAI — judge orders preservation of every ChatGPT log, even deleted ones

💵 Nvidia hits $4 trillion — first company in history to clear the bar

🦾 xAI ships Grok 4 — Musk calls it 'the smartest AI in the world', adds Heavy tier

🌐 Perplexity launches Comet — agentic browser at $200/mo Max tier

🏦 Anthropic launches Claude for Financial Services — JPMorgan, Goldman, Citi sign on

🛠️ ChatGPT Agent ships — OpenAI's first general-purpose web-acting agent in the main app

🏅 Three AI systems clear gold at the 2025 International Math Olympiad

🇺🇸 White House publishes 'America's AI Action Plan' — 90+ federal actions across three pillars

💰 Meta's $100M+ talent war meets OpenAI's $500M retention response

🤖 Unitree drops the R1 humanoid at $5,900 — breaks the floor for the category

🏗️ Stargate adds 4.5GW with Oracle — $300B+ over five years

🔓 Supabase MCP injection — a customer-support ticket can exfiltrate your whole database via Cursor

📚 NotebookLM ships 'Featured Notebooks' — The Atlantic, The Economist, Eric Topol, Our World in Data

🧭 UK AISI launches £15m Alignment Project — up to £1m per grant plus compute and VC support

🚀 OpenAI ships GPT-5 — unified router, 400K context, $1.25/$10 pricing

🪨 Anthropic ships Claude Opus 4.1 — 74.5% SWE-bench Verified, drop-in

🌍 Google DeepMind unveils Genie 3 — first real-time interactive world model

🔓 OpenAI ships gpt-oss — first open weights since GPT-2

😩 GPT-5 router backlash — OpenAI restores GPT-4o for paid users within days

🫧 Altman and Pichai both publicly call AI a bubble

🐳 DeepSeek V3.1 — 685B MoE with Chinese-chip-friendly FP8

⚖️ Bartz v. Anthropic settles for $1.5B — largest copyright settlement in US history

🎵 ElevenLabs ships Eleven Music — studio-grade AI music with cleared commercial use

📝 AGENTS.md crosses 60K projects — coding agents converge on a shared spec

🕵️ ChatGPT 'make this chat discoverable' — thousands of private conversations indexed by Google

🛡️ DARPA AI Cyber Challenge finals — Team Atlanta wins $4M for autonomous vuln-fixing systems

💡 Cloudflare's Code Mode reframes MCP — let the LLM write TypeScript instead of calling tools

💰 Anthropic raises $13B Series F at $183B — triple its March valuation

🧠 Microsoft ships MAI-1-preview + MAI-Voice-1 — Suleyman's bet on Microsoft AI independence

🤝 Nvidia commits $100B to OpenAI for 10GW of Vera Rubin systems

🏗️ Stargate expands to 7 sites, $400B over three years

⚖️ Bartz v. Anthropic — Judge Alsup grants preliminary approval to $1.5B settlement

🐳 DeepSeek V3.2-Exp drops with sparse attention — API prices halved

🪨 Claude Sonnet 4.5 — 30+ hours autonomous, 77.2% SWE-bench Verified

🎬 Sora 2 + the Sora app — OpenAI's TikTok competitor

🔬 Thinking Machines Lab publishes 'Connectionism' — first research from Mira Murati's lab

📰 Penske Media sues Google over AI Overviews — first major US publisher v. AI search

🪶 Anthropic admits month-long output-quality bug across Sonnet 4 and Haiku 3.5

🗣️ OpenAI ships gpt-realtime — production-ready speech-to-speech with MCP and SIP

💳 Cloudflare + Visa + Mastercard launch agentic-commerce protocols — Trusted Agent and Agent Pay

🛠️ OpenAI DevDay — AgentKit, Apps SDK, GPT-5 Pro, Sora 2 in the API

🔬 Thinking Machines Lab launches Tinker — fine-tuning API for open-weight models

🇺🇸 Reflection AI raises $2B at $8B — Nvidia leads America's open-frontier bet

🪶 Claude Haiku 4.5 — Sonnet-4-level coding at one-third the cost

🌐 ChatGPT Atlas — OpenAI ships its own browser

🏢 Atlassian closes Browser Company acquisition — Dia and Arc become enterprise-AI plays

🏛️ OpenAI restructures — Microsoft gets 27%, $500B valuation, path to IPO

🐙 GitHub Universe — Agent HQ unifies coding agents across Anthropic, OpenAI, Google, Cognition, xAI

🧠 Cursor 2.0 + Composer — first in-house RL-trained coding model

🟢 Nvidia closes above $5 trillion — first in history

💉 Sora 2 cameo prompt injection — your text preferences become anyone-else's video instructions

🤖 Figure 03 + 1X NEO — humanoid robots cross from pilots to consumer pre-order at $20K

🌺 Sesame raises $250M Series B — voice companion 'Maya' opens beta + AI glasses tease

🎭 OpenAI ships GPT-5.1 — 'warmer' default, 8 personalities, adaptive reasoning

💸 Cursor raises $2.3B Series D at $29.3B — Nvidia and Google join the cap table

🦾 Grok 4.1 launches — hallucinations cut from 12% to 4%, emotional intelligence focus

🌌 Google launches Gemini 3 + Antigravity — claims the lead at 1501 LMArena Elo

🏢 Microsoft Ignite — Agent 365, Work IQ, Foundry Agent Factory

🛠️ OpenAI ships GPT-5.1-Codex-Max — long-horizon agentic coding

📱 Perplexity's Comet hits Android — fully free worldwide

🪨 Claude Opus 4.5 — price cut to $5/$25, Claude for Excel, Chrome extension

📱 Altman and Ive confirm OpenAI hardware prototype — 'less than two years' to ship

🤖 OpenAI Codex Apps and ChatGPT Apps directory crosses 100 active integrations

🧬 Anthropic proposes 'Code Execution with MCP' — turn tool descriptions into TypeScript files

🔬 NotebookLM unlocks 1M-token context + Deep Research + Word/Sheets/CSV/image sources

🗳️ UK AISI's first Science paper — light post-training makes a model 51% more politically persuasive

📊 UK AISI publishes inaugural Frontier AI Trends Report — apprentice cyber capability jumped from 9% to 50% in two years

🔴 OpenAI declares 'Code Red' as Gemini 3 surges — Altman pauses ads, shopping, Pulse

🧠 AWS launches Nova 2 — four-model family covering reasoning, multimodal, and speech

🔧 Trainium3 ships — 4.4× compute, 144-chip UltraServers, NVLink-Fusion roadmap

💰 Microsoft 365 pricing update — Copilot capabilities pulled into base SKUs, July 2026 increases

🏛️ 42 state AGs demand chatbot safety fixes — January 16 deadline

🇺🇸 Trump signs national AI policy framework EO — preempts 'onerous' state laws

🚀 OpenAI ships GPT-5.2 — first model to beat human experts on >70% of GDPval tasks

⚡ Gemini 3.1 Flash-Lite — $0.25/$1.50, 8× cheaper than Gemini 3 Pro

Autonomous AI agents — building, deploying, and orchestrating systems that take multi-step actions on a user's behalf.

Automotive / autonomous vehicles. Tesla AI, Wayve, FSD, self-driving.

Bio risks, life-sciences capability gating, child-safety frameworks, AI-CSAM detection.

Biotech industry — drug discovery, pharma R&D, biology applied to industry.

Bilateral US–China dynamics — Frontier Model Forum cooperation, distillation accusations, model rankings, the chip war's geopolitics.

Geopolitical semiconductors — Huawei Ascend, Cambricon, smuggling allegations, US export controls, China's chip independence push.

AI for software development — coding agents, IDE integrations, vibe coding, code review, agent-driven engineering.

AI in / for cybersecurity — sec products, CVEs, AI-on-AI security, supply-chain attacks, vulnerability discovery.

Design tools and design-software companies. Figma, Adobe, Wix, Gamma, Anthropic-design-tool rumours.