Gemini takes on Chrome tasks as OpenAI discusses $100 billion funding

Google announced “auto browse” for Gemini in Chrome on January 28. The feature is rolling out gradually as a preview for Google AI Pro and Google AI Ultra subscribers using desktop Chrome in the US.¹² Gemini can interact with websites by clicking, scrolling, and filling out forms. Google says it can help with multistep tasks such as comparing products and booking travel.¹ Google requires confirmation before purchases and other sensitive actions.¹

The feature brings an AI browser agent into Chrome rather than requiring users to switch to a separate AI browser. It draws on Google’s earlier Project Mariner work.³ Access remains limited and the rollout is gradual; Google has not given a date for wider availability.¹² Businesses should also note that Google’s help page says auto browse is unavailable with work or school accounts.²

OpenAI is discussing a funding round of up to $100 billion, according to media reports. NVIDIA, Microsoft, and Amazon are among the potential investors. The talks became public in late January, but no completed round had been confirmed by early February. SoftBank and Middle Eastern sovereign wealth funds were also identified as potential backers. How much each participant would ultimately invest remained uncertain.¹²

The proposed round reflects the capital needed to expand AI data centers. In September 2025, NVIDIA announced a letter of intent to invest up to $100 billion to support new infrastructure with OpenAI. That plan later stalled, and the amount was not a firm commitment. A smaller potential NVIDIA contribution to the current round was reported in early February, also without a confirmed deal. The figures under discussion should not be treated as committed funding.³⁴⁵

Apple confirmed its acquisition of Israeli AI startup Q.ai on January 29, 2026. The company works on audio processing and machine learning. Reports describe technology intended to detect whispered or silent speech through subtle facial movements. Q.ai CEO Aviad Maizels and co-founders Yonatan Wexler and Avi Barliya are joining Apple.¹²

Apple disclosed neither a price nor specific product plans. One report put the deal at nearly $2 billion, while Reuters cited a lower estimate.¹³ The technology could be relevant to headphones, smart glasses, or AI assistants. Apple has not announced such a feature or a launch date, however.³² If the reported price is accurate, Q.ai would rank among Apple's largest acquisitions.³²

Alibaba’s Qwen team open-sourced Qwen3-ASR and Qwen3-ForcedAligner in late January. The two ASR models recognize speech across 52 languages and dialects and support both recorded audio and streaming input. Qwen says they can also process singing and songs with background music. Qwen3-ForcedAligner adds word- and phrase-level timestamps to transcripts.¹²

The release includes inference and fine-tuning tools, with vLLM support for batch processing, streaming, and asynchronous serving. That gives teams options for transcribing stored conversations or processing audio as it arrives. Companies should still test accuracy on their own accents, background noise, and specialized vocabulary: the stated capabilities come from Qwen, not from an independent evaluation described in these sources.¹²

xAI introduced the Grok Imagine API on January 28, 2026. Developers can use it to add video generation and editing to their applications. The API supports video creation from text and images, as well as changes to existing scenes. According to xAI, users can add or remove objects and control motion.¹² On February 2, the company announced Grok Imagine 1.0 with details on video length, resolution, and improved audio.³

For companies, the combination of generation and editing could make it easier to create and revise footage within one workflow. The described editing capabilities come from xAI.¹ The announcement alone does not establish how reliably they work on complex source material. Teams considering the API for production use should test it with their own scenes and requirements before building workflows around it.

On January 29, 2026, Anthropic published a controlled study of how AI assistance affects learning a new Python library.¹ Participants coded with or without AI help, then answered questions about the code they had written.¹ The AI group averaged 50% on the follow-up comprehension quiz, compared with 67% for the group working without AI.¹ The difference was especially pronounced on debugging and error-diagnosis questions.¹

The result concerns comprehension immediately after a learning task; it does not establish an effect on long-term retention.¹ AI assistance also showed no clear overall speed advantage on this unfamiliar task.² For companies, the distinction between completing work and understanding it matters: finishing code with AI help does not necessarily mean a developer can explain it or diagnose errors independently.¹

Hugging Face introduced Daggr on January 29, 2026, as an open-source Python library for AI workflows.¹ Developers can use it to connect Gradio apps, ML models, and custom functions in code.¹ Daggr automatically generates a visual canvas where users can inspect intermediate outputs and rerun individual steps.¹

The library combines code-based workflow construction with a visual interface for debugging.¹ Its integrations focus on Gradio and Hugging Face services.¹ Teams considering production use should account for its maturity: the package documentation says Daggr is in active development and its APIs may change.² Updates also carry a risk of losing locally persisted state.²