Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
Anthropic filed a confidential S-1 with the SEC, setting up what could be the largest AI company IPO in history — ahead of OpenAI's own Q4 target. The filing follows a $65B Series H and cites a $47B annual revenue run-rate. Anthropic would become the first frontier model lab to trade publicly.
Microsoft unveiled Project Polaris at Build 2026, a homegrown mixture-of-experts model that replaces GPT-4 Turbo as GitHub Copilot's default engine by August 2026. Polaris outperforms GPT-4 Turbo on HumanEval and MBPP with strong gains in Rust and Haskell. This is Microsoft claiming full-stack ownership of its most-used developer product after three years of OpenAI dependency.
Also at Build, Microsoft AI chief Mustafa Suleyman introduced MAI-Thinking-1, Microsoft's first frontier reasoning model, in private preview via Azure AI Foundry. The model targets high efficiency at low token cost and marks Microsoft's intent to compete at the frontier model layer — not just resell OpenAI output.
Anthropic released Claude Mythos Preview to select companies under Project Glasswing, a cybersecurity AI program. The model has reportedly found thousands of high-severity vulnerabilities in major operating systems and browsers on its own. Anthropic is actively briefing the Trump administration on its capabilities.
Gemini 3.5 Flash went GA today at four times the speed of comparable models, priced at $1.50 per million input tokens with a one-million token context window. It scores 76.2% on Terminal-Bench 2.1, beating Gemini 3.1 Pro on coding and agent tasks, and becomes the default model in the Gemini app and Google AI Search Mode.
Anthropic introduced a Managed Agents layer: a production execution environment separating agent logic from runtime concerns — orchestration, sandboxing, state management, credentials. The move positions Claude as a complete enterprise agent platform and puts it in direct competition with Microsoft's Azure Agent Mesh, announced the same morning.
OpenAI co-founder and former Tesla AI lead Andrej Karpathy announced he is joining Anthropic to work on pre-training under Nick Joseph, helping build a new research team that uses Claude to accelerate the work. He posted that "the next few years at the frontier of LLMs will be especially formative." The hire lands days before Anthropic's IPO filing.
OpenAI's frontier models and Codex are now accessible directly through AWS infrastructure, a meaningful multi-cloud move that challenges the perception of OpenAI as an Azure-exclusive partner. Enterprise teams already running on AWS can now deploy OpenAI models without routing through Microsoft's cloud.
DeepSeek's promotional 75% price reduction on V4-Pro is now the permanent list price at roughly $0.87 per million output tokens. V4-Pro scores 80.6 on SWE-bench Verified versus Claude Opus 4.7's 80.8 — near-identical performance at a fraction of the cost — and DeepSeek says the cuts stem from genuine architectural efficiency, not subsidization.
Microsoft rounded out Build with Azure Agent Mesh for federated multi-agent execution across clouds and devices, and released Windows Agent Framework 1.0 under the MIT license. Satya Nadella framed the shift as AI moving from synchronous assistant to async coworker running long-horizon tasks. GitHub Copilot Workspace also went GA for enterprise.
