Last Week’s AI News #4

Last week was particularly intense in the AI world – there was no shortage of news! Artificial intelligence continues to transform business, technology, and science, opening opportunities for innovation, optimizing processes, and providing a competitive advantage. In this blog, we review the most important events that may be relevant to both large enterprises and small-to-medium businesses (SMBs).

GOOGLE LAUNCHES GEMINI ENTERPRISE: “ALL-IN-ONE” AI PLATFORM

Google introduced Gemini Enterprise, a platform that allows users with no programming experience to create, deploy, and manage AI agents in a single environment. It combines no-code development tools with ready-made assistants (for research, coding, customer support), securely connects to company data, and offers a marketplace with thousands of partner solutions. The Enterprise plan costs $30/month per user, while the cheaper Business plan is $21/month.

Why this matters for businesses:

• Process efficiency: Automates daily tasks, reduces labor costs, and increases productivity.

• Faster innovation deployment: No-code tools enable teams to quickly create and adapt AI solutions to business needs.

• Strategic advantage: Direct AI integration into workflows accelerates decision-making and enhances competitiveness.

AI ADOPTION GROWS, BUT TRUST REMAINS LOW

A Reuters Institute survey in six countries showed that weekly AI usage nearly doubled over the past year and is expanding in scope, yet the public remains skeptical about AI in news content.

• Main uses of AI: Information search – 24% of respondents use AI for research and queries, 21% for generating text, images, or code.

• ChatGPT remains the clear leader in usage, while Google and Microsoft integrated search engines allow 54% of users to see AI-generated summaries.

• Only 12% feel comfortable reading AI-generated news, and 62% prefer human-written content – the trust gap has widened since 2024.

• AI is viewed most positively in healthcare, science, and research, and negatively in media and politics.

Why this matters for businesses:

• Innovation potential: Growing AI usage opens new opportunities to optimize processes.

• Human-AI balance: Combining automation with human judgment strengthens productivity and user trust.

AI RESEARCHER LEAVES ANTHROPIC OVER CHINA POLICY

Renowned physicist turned AI researcher Yao Shunyu left Anthropic in less than a year and joined Google, citing the startup’s stance toward China as one of the reasons.

• Yao contributed to Claude 3.7 Sonnet and Claude 4 development.

• He stated that 40% of his decision was influenced by Anthropic’s policy restricting subsidiaries in “hostile nations.”

• DeepMind recruited Yao to the Gemini team, where he will work on foundational models.

Why this matters for businesses:

• Project continuity: International restrictions can impact AI project implementation.

• Strategic competitive advantage: Companies aware of geopolitics can better maintain the pace of AI innovation.

SAMSUNG TRM: COMPACT AI MODEL OUTPERFORMS GIANTS

Samsung researcher Alexia Jolicoeur-Martineau introduced the Tiny Recursion Model (TRM) – a 7-million-parameter AI model that outperforms DeepSeek R1 and Gemini 2.5 Pro in complex reasoning tasks using a self-improvement cycle that includes solution creation, reflection, and refinement.

• TRM achieved 45% on ARC-AGI-1 and 8% on ARC-AGI-2, outperforming much larger models.

• Instead of step-by-step answer generation, TRM creates solutions and iteratively refines them through up to 16 internal reasoning and review cycles.

• The model uses a separate “notepad,” where each cycle critiques and improves its logic six times before updating answers.

• Results were especially promising for specific puzzle-type tasks in ARC tests but may not generalize across all reasoning domains.

Why this matters for businesses:

• Efficiency and accuracy: TRM shows that even smaller AI models can solve complex problems efficiently, saving resources and time.

• Innovation potential: Self-improvement cycles and iterative solution refinement can be applied to business analysis, optimization, and decision-making.

• Competitive advantage: Companies implementing advanced, even compact, AI models can achieve faster effective solutions and market advantage.

OPENAI DEV DAY 2025: CHATGPT UPDATES, AGENT CREATION TOOLS, AND ADVANCED MODELS FOR DEVELOPERS

OpenAI held Dev Day 2025 last week, presenting new app integrations directly within the ChatGPT platform, agent creation capabilities, expanded access to models like Sora 2 via API, and more.

• Users can now run, interact with, and build apps directly in ChatGPT using the Apps SDK.

• Apps will open and integrate directly into the ChatGPT chat interface, with monetization options.

• AgentKit – a new toolkit for building agents.

• GPT-5-Codex is now available to all, with Slack integration, SDK customization for workflows, and enterprise management.

• Developers also get API access to GPT-5 Pro for advanced reasoning, Sora 2, and the gpt-realtime-mini voice model, which is 70% cheaper than previous versions.

Why this matters for businesses:

• Faster innovation: New app integrations and agent creation tools enable companies to deploy AI solutions without complex programming infrastructure.

• More efficient workflows: Using models directly in ChatGPT and integrating them into business processes reduces time and increases productivity.

• Cost optimization: Cheaper voice and reasoning models allow businesses to use advanced AI at lower costs.

• Monetization and integration opportunities: Companies can create products and services integrated with ChatGPT, expanding business model potential.

Read more about Open Dev Day.

GOOGLE GEMINI 2.5 COMPUTER USE

Google introduced the preview of Gemini 2.5 Computer Use – a new AI model accessible via API that can control web browsers and perform tasks by directly interacting with the user interface, e.g., clicking buttons or filling out forms.

• The model works by capturing website screenshots and analyzing them to autonomously execute clicks, typing, and navigation commands.

• Gemini 2.5 Computer Use outperformed competitors, including OpenAI’s Computer Using Agent and Claude Sonnet 4.5/4, in both web and mobile tests.

Why this matters for businesses:

• Automation and efficiency: Gemini 2.5 can perform repetitive tasks directly in the browser, saving employee time and reducing errors.

• Faster process execution: The model operates with low latency, completing tasks faster than competitor solutions.

• Integration with business systems: API access allows businesses to easily incorporate AI into existing apps and tools.

IBM NETWORK INTELLIGENCE

IBM Network Intelligence introduced a new human-AI collaboration model that combines analytical and reasoning AI to help companies filter noise, identify problems, and scale operations smoothly.

Why this matters for businesses:

• Faster decision-making: Combining analytical and reasoning AI enables companies to quickly identify problems and obtain accurate insights.

• More efficient operations: Fewer redundant tools mean higher efficiency and lower costs.

• Strategy management: AI processes data automatically, letting humans focus on decision-making and strategy.

DUKE UNIVERSITY AI SYSTEM FOR MORE EFFICIENT DRUG DELIVERY

Duke University researchers introduced TuNa-AI – a platform combining robotics and machine learning to create nanoparticles for drug delivery, significantly improving cancer treatment efficiency.

• 1,275 formulas tested, achieving 43% higher successful nanoparticle creation than traditional methods.

• Successfully created a hard-to-deliver leukemia drug with better solubility and higher cancer cell kill rate.

• Reduced potentially toxic substances by 75% while maintaining treatment effectiveness.

• TuNa simultaneously manages material selection and mixture ratios, overcoming limitations of existing methods.

Why this matters for businesses:

• Innovation advantage: Companies using such solutions can deploy new treatment methods faster and experiment more productively.

• Safety and cost optimization: Reducing toxic components maintains effectiveness while lowering production risks and costs.

POPULAR AI TOOLS

• Incogni – Helps remove personal data online and protects against fraud and identity theft.

• Gemini Enterprise – AI agents in one platform.

• Amazon Quick Suite – Quick information retrieval across platforms.

• ElevenLabs UI – Open-source components for AI-based voice and audio agents.

• CData Connect AI – Connects any data source to AI for real-time company insights.

• Grok Imagine v.0.9 – Updated xAI image and video generation platform.

• Hunyuan-Vision-1.5-Thinking – Advanced Tencent vision and language model.

• Petri – Open-source agent tool from Anthropic for LLM security evaluation.

• Sora 2 – Advanced OpenAI video model, now available via API.

• Code – OpenAI agent programming tool, now accessible to all.

OTHER AI DEVELOPMENTS LAST WEEK

• Amazon Quick Suite – Agent platform for connecting information across apps, automating processes, and performing research.

• Zendesk new AI agents – Can resolve up to 80% of customer queries with voice and co-pilot agents.

• OpenAI ChatGPT reached 800M weekly users – API now processes over 6B tokens per minute, enabling SMB integration.

• Google CodeMender – AI agent automatically finds and fixes software bugs for SMB system security.

• Adobe new B2B marketing agents – Audience, Journey, and Data Insights tools for automating marketing processes.

• ElevenLabs Agent Workflows – Visual tool for creating interactive voice interactions.

• Sora 2 video model – Fast, accessible AI video tool for SMB content creation.

• Google AI try-on and footwear features – Allows small e-commerce businesses to showcase products individually, boosting sales.

• ASAPP Generative AI Agent 100 – 100 AI agent use ideas for contact centers, reducing costs and accelerating services.

Last week showed that AI tools and models are rapidly establishing themselves in business, science, and healthcare. This enables increased efficiency, cost reduction, faster solution development, and maintaining competitiveness.

Subscribe to our newsletter to receive the most important AI news directly in your inbox every week!

Don't miss news