TLDR AI 2025-11-19

The Manus Browser Operator is a browser extension that allows Manus to operate within users' local browsers. It is currently rolling out in beta to Pro, Plus, and Team users. The extension works on Chrome and Edge - wider browser compatibility has been planned. Users maintain full control and transparency over every action. A short video demonstrating its functions is available in the article.

Gemini 3 achieves state-of-the-art performance on most major benchmarks, including a breakthrough 1501 Elo on LMArena, 91.9% on GPQA Diamond, and 81% on MMMU-Pro. The model includes a Deep Think mode for enhanced reasoning and coincides with the launch of a new agentic coding IDE called Antigravity.

Anthropic has signed new strategic partnerships with Microsoft and Nvidia. Anthropic has committed to purchasing $30 billion of Azure compute capacity and contracted for additional compute capacity up to 1 gigawatt. It has also committed to purchasing up to 1 gigawatt of compute capacity with Nvidia's Grace Blackwell and Vera Rubin systems. Microsoft will invest up to $5 billion into Anthropic, while Nvidia will invest up to $10 billion.

AI is moving fast - every week brings a new agent framework, evaluation suite, orchestration layer, or open-source model. The AI in Practice Survey looks at what teams are actually adopting in production and what problems nobody has solved yet. It maps where adoption is happening, where the gaps are, and how teams are hardening and scaling AI in practice. The results of the survey have been compiled into an interactive dataset.

Gemini 3 feels more consistent and has less 'spike' than previous models. It doesn't output 'AI slop' on creative writing tasks anymore. The model is fast, and it often outperforms GPT-5 Pro without the wait. The Antigravity IDE is powerful, but requires active supervision to catch errors.

AI companies are diversifying beyond Nvidia chips due to lead times, costs, and concentration risks. The market is seeing signals of a multi-chip landscape, as software improves and adoption of non-Nvidia solutions increases, evidenced by hiring trends and strategic partnerships like OpenAI's alignment with AMD and Broadcom. This diversity is expected to reduce lead-time risk and improve key performance metrics for startups.

Stop juggling vendors. AssemblyAI combines multilingual Speech-to-Text, Speech Understanding, speaker diarization, and LLM integrations - all in one developer-friendly API. It's trusted by voice AI leaders like Granola, Dovetail, and Ashby. It's free to try, then pay-as-you-go with no upfront commitment. Just get your API key and start building.

Antigravity is a free VSCode-forked IDE that runs multiple autonomous AI agents with direct access to editors, terminals, and browsers, generating "Artifacts" like task lists, screenshots, and browser recordings to document their work.

Parallel's FindAll API turns natural language queries into custom datasets. It finds entities like companies, people, or locations based on search criteria, then enriches them with structured data along with citations. The tool is extensible and can produce high-quality datasets on demand. It excels at entity discovery and research tasks that require both breadth and depth.

Stack Overflow has revealed a new set of products aimed at positioning the company as a valuable part of the enterprise AI stack. The company is attempting to remake its classic problem-solving forum into a tool for translating human expertise into an AI-accessible format. Slack Internal is an enterprise version of the web forum that exports a layer of metadata alongside question and answer pairs. It includes a writing function that allows agents to create their own queries if they can't answer a question or notice a knowledge gap.

Modern GPUs threaten to accelerate asset depreciation beyond what even some of the largest companies can handle. Analysts worry the rapid pace of AI processing power advances could overwhelm companies riding the AI train. New advances quickly make data centers running last-generation hardware decidedly unprofitable. Everyone is competing for a limited pool of hardware, and no one is buying older GPUs, making operations even harder to sustain. With spiking electricity costs, increasing pressure for eco-conscious data center design, and a lack of clear profitability for AI, it's a disaster waiting to happen.

28% of all business calls go unanswered. CallRail's Voice Assist is a human-like AI receptionist that answers the phone, helps the caller, and ties every call to its marketing source. Hear how it sounds on a call ☎️

Anthropic's Claude Sonnet 4.5, Haiku 4.5, and Opus 4.1 are now available in public preview through Microsoft Foundry and Microsoft 365 Copilot, allowing Azure users to build with Claude's capabilities for coding, agents, and productivity tasks.

Soumith Chintala, one of the creators of PyTorch, is joining Thinking Machines Lab, just weeks after leaving an 11-year run at Meta.

Azure's new ND GB300 v6 VMs, using NVIDIA's Blackwell architecture, delivered 1.1 million tokens per second in AI inference, setting a new MLPerf Inference record.

Investors aren't happy that Oracle is pouring every last drop of capital into the OpenAI pipe dream with the hopes that it will pay off later.