This is Hamamoto from TIMEWELL.
The pace of AI development has reached a level where meaningful updates arrive weekly. This article covers three major tool developments — Gemini's Canvas mode evolution, Kimi K2 Thinking's benchmark results, and SkyReels' generative media capabilities — alongside the industry news that frames where this technology is heading: investment at unprecedented scale, legal clarity beginning to emerge, and infrastructure investment accelerating globally.
Part 1: Tools — What's New and What It Does
Gemini Canvas: Presentations from a Single Prompt
Gemini's Canvas mode now generates complete presentation slides from a single text prompt. The workflow:
- Log in with a Google account (free)
- Upload or reference a PDF document
- Prompt: "Create a presentation based on this document"
- Gemini generates slides; export directly to Google Slides for editing
The Canvas mode enables editing text and code efficiently, and the Google Slides export allows further customization. The process that previously required multiple steps — extracting content, formatting slides, adjusting layout — now happens in a single generation cycle.
Deep Research integration: Gemini's Deep Research function has been updated to connect with Gmail, Google Drive, and Google Chat, functioning as a research agent across these data sources. For teams already operating within Google Workspace, this makes Gemini a practical research layer across internal information.
Kimi K2 Thinking: Open-Source Reasoning Benchmarks
Moonshot AI released Kimi K2 Thinking, an open-source reasoning model that performed at the top of several benchmarks:
- Top scores on highly difficult reasoning benchmark tests
- Top performance on BrowseComp (evaluation of AI web search capability)
- Performance comparable to GPT-5 on multiple evaluation criteria
The model is accessible for free through the Moonshot AI official chat. Select "K2" from the tab in the lower right of the interface, then toggle "Thinking" on. Users have reported the model navigating complex logic problems with notable speed and flexibility — handling cases where previous AI systems would require more iterations.
SkyReels and the Generative Media Landscape
SkyReels added AI chat-based image and video generation — enabling content creation through natural language interaction. Combined with other platforms in this space (Bing MAI-Image-1, Grok's text-to-video, Canva's expanded AI features), the generative media toolset available across different subscription tiers has become substantially broader.
Canva AI expansion: Canva implemented a large batch of AI features, bringing generation and editing capabilities further into its standard creative workflow.
OpenAI Android video app: OpenAI released a dedicated video generation application for Android, extending Sora's capabilities to mobile users.
Looking for AI training and consulting?
Learn about WARP training programs and consulting services in our materials.
Part 2: Industry Moves — Investment, Legal, and Infrastructure
OpenAI + Amazon: ¥5.8 Trillion Cloud Deal
OpenAI signed a seven-year cloud service agreement with Amazon, valued at approximately ¥5.8 trillion ($40 billion). Previously, Microsoft was the primary cloud partner for ChatGPT's infrastructure. The Amazon deal diversifies the infrastructure dependencies and provides stable compute capacity for ChatGPT development and operation.
The scale of this contract — a seven-year, multi-trillion yen commitment — reflects the infrastructure cost reality of large-scale AI deployment. Cloud infrastructure is not a marginal cost; it is a primary structural expense for frontier AI companies.
SoftBank + OpenAI Japan Joint Venture
SoftBank Group and OpenAI established a Japan-focused joint venture ("SBOAIJAPAN LLC") to bring OpenAI's products to Japanese enterprise customers with localized implementation support and operational guidance. Japan's government has also committed to prioritizing investment in AI, robotics, and 17 other designated growth sectors, with a new growth strategy targeting the following year.
The public-private alignment in Japan's AI push represents a structured approach to building domestic AI capability and adoption infrastructure.
Getty Images vs. Stability AI: UK Copyright Ruling
The UK court ruled that Stable Diffusion's training process on existing images did not constitute copyright infringement under UK law. This ruling provides some legal clarity on a question that has been unresolved: whether training an AI model on copyrighted material constitutes infringement, independent of what that model subsequently generates.
The ruling is expected to influence AI copyright policy discussions across jurisdictions. For AI companies building on training data from the web, it represents a significant data point in an ongoing legal environment.
Additional Industry Context
- Apple: Added live translation to iOS 18.1 and iPadOS 18.1, enabling real-time text and voice translation; reports of potential adoption of a Google-developed large AI model for Siri
- AMD: Q4 AI product sales exceeded analyst expectations, reflecting continued data center investment growth
- Google "Project SunCatcher": Reported plans for space-based AI data center infrastructure — a longer-horizon infrastructure concept
- Perplexity browser "Comet": Received warnings from Amazon regarding its ability to make purchases on Amazon on behalf of users — a signal of emerging friction between AI agents and platform access policies
- OpenAI CFO on IPO: Signaled no near-term IPO plans while the company's valuation approaches $1 trillion
- Cultural penetration: Multiple AI-related terms nominated for Japan's 2025 New Word Award, reflecting how broadly AI has entered everyday vocabulary
Part 3: What These Changes Mean for Work and Business
AI as Decision Support Infrastructure
Models like Gemini Deep Research and K2 Thinking don't just answer questions — they synthesize large volumes of information to provide analysis oriented toward the user's specific goal. Work that previously required commissioning specialists for research and analysis is becoming accessible at lower cost and faster timelines.
This changes the economics of information-intensive work. Individuals and small organizations can access analytical capability that was previously available only to large teams.
From Tool to Partner
The shift in generative AI's role in creative and knowledge work is from task executor to collaborative partner. Video generation, image generation, presentation creation, and research synthesis are increasingly functions where AI handles structure and initial execution while humans direct the intent and evaluate the output.
The practical effect: less time spent on execution, more time available for judgment and direction.
Infrastructure Is Hardening
The OpenAI-Amazon deal, Google's data center investments, and government initiatives across multiple countries signal that AI infrastructure is being treated as strategic infrastructure — not just commercial tooling. The reliability and availability of AI services is being built to a higher standard.
For businesses relying on AI tools, this means the underlying infrastructure is becoming more stable and more capable simultaneously.
Legal Clarity is Emerging
The Getty-Stability ruling is one data point, but it points toward a broader trend: courts and regulators are beginning to define the rules around AI training data, copyright, and platform access. This clarity — even when unfavorable to specific parties — reduces the uncertainty that has surrounded AI product development.
Summary
Key developments covered:
- Gemini Canvas: Single-prompt presentation generation with Google Slides export; Deep Research now integrates Gmail, Drive, and Chat
- Kimi K2 Thinking: Open-source reasoning model with top benchmark scores comparable to GPT-5; free to use via Moonshot AI chat
- SkyReels and generative media: Expanded image and video generation through chat interfaces across multiple platforms
- OpenAI + Amazon: ¥5.8 trillion, seven-year cloud infrastructure deal
- SoftBank + OpenAI Japan JV: Localized enterprise deployment for Japanese market
- Getty vs. Stability AI: UK court rules AI training did not infringe copyright
- Broader trend: AI shifting from business efficiency tool to foundational infrastructure
The gap between organizations that integrate AI into their core workflows and those that don't is widening. Understanding what these tools do — and how the environment around them is evolving — is increasingly a baseline business competency.
Reference: https://www.youtube.com/watch?v=3Rukvd9olSA