The most important news pieces from the AI world are here for you, whether you’re an AI enthusiast or just wish to keep an eye on what’s going on.
Focus Topic: OpenAI Dev Day 2025: the Agent Shift
At this year’s OpenAI Dev Day, the main focus wasn’t on launching new models, but on building out the infrastructure for how AI is expected to be used in everyday tasks. The standout announcements centered around tools that help move AI from passive assistant to active participant.
The most notable of these was AgentKit – a toolset for building AI agents that can complete multi-step workflows on their own. This includes visual builders, human-in-the-loop components, and integrations with external systems. The idea is to simplify what used to require separate tools like Zapier or Make. These agents can now be built and run entirely within the ChatGPT environment.
Alongside that came the Apps SDK, which allows developers to create mini-apps that run inside ChatGPT. Apps like Canva, Coursera, and Zillow were shown working seamlessly within the chat. Users can interact with them through natural conversation – generating content, getting information, or completing tasks without switching platforms. You don’t leave the interface to use them. You just ask: “Canva, turn this into a poster,” and it happens.
The third major update was a production-ready version of Codex, OpenAI’s code-writing assistant. Now based on the GPT-5 architecture, it supports more complex programming tasks, including API integration, hardware control, and voice-based commands. In the live demo, Codex connected to a Sony camera and Xbox controller, built UI elements, and even edited running code – all via natural language instructions.
Together, these tools lower the barrier to building functional software. They don’t remove the need for design or thought, but they change who can do the building, and how.
LLMs & AI Models
- Google launched Gemini Enterprise, a unified AI platform for building and deploying custom workplace agents with no code.
- Claude Code added plugin support, enabling custom workflows, internal tools, and collaborative coding via simple plugin commands
- Samsung introduced the Tiny Recursion Model (7M), which outperformed larger models in logic tasks using multi-step internal evaluation cycles.
- Google previewed Gemini 2.5 Computer Use, an API model that can autonomously control browsers by analyzing screen images.
- ChatGPT now has 800 million weekly active users, and OpenAI’s API processes over 6 billion tokens per minute.
- Deloitte Australia refunded part of its AI-generated report fee after using Azure OpenAI, which produced fabricated citations.
- Microsoft improved Copilot by adding direct document creation, email integration, and one-click export for long responses.
New Tools
- Jony Ive and Sam Altman are designing a new family of AI devices, aimed at creating calmer, more human-centered tech experiences.
- Google’s virtual try-on now includes footwear and expands to Australia, Canada, and Japan, with full-body photo input support.
- xAI launched Grok Imagine v0.9, offering improved video quality, synced audio, and camera effects.
- Google introduced PASTA, an AI that adapts to user aesthetic preferences for image generation without prompt engineering.
👉 Explore these tools: Grok Imagine | Google PASTA | Gemini Try-On
Other Quick Picks
- A Reuters study shows AI use nearly doubled in a year, but only 12% fully trust AI-generated newscompared to 62% who prefer human-written content.
- Sam Altman said future companies may operate with zero employees, as AI agents can complete week-long tasks independently.
- OpenAI and AMD signed a deal for 6 GW of compute power, giving OpenAI a future option to acquire 10% of AMD shares.
- Analysts warn that AI funding loops between OpenAI and Nvidia may inflate revenue, raising concerns about sustainability.
- OpenAI’s Sora app hit 1 million downloads in under five days, faster than ChatGPT, despite invite-only access.
- MrBeast warned that high-quality AI video generation could threaten millions of creators, raising copyright concerns.
- Yao Shunyu, former Anthropic researcher, left over anti-China policy, now working at DeepMind on Gemini models.
🇪🇪 AI News from Estonia
- I was on the stage of World Summit AI , twice! And it was amazing. I did a keynote to full house and shared how AI and CEO’s must go hand-in-hand, otherwise there will be a big problem with the legal side. And I moderated also a panel about the future of banking and I haven’t heard sweeter feedback before 🙂
- Allan Martinson told the Estonian parliament that AI tools could help raise productivity to hundreds of thousands of euros per employee.
- Estonia’s Ministry of Justice plans to replace free legal advice with a chatbot by 2026; Hugo.legal will offer a paid alternative.
- PPA chief Egert Belitšev said AI could help detect falls and public safety incidents via live camera feeds before humans notice them.
- Estonian startup Dragonfly raised €3M to launch an AI-based software recommender for companies, analysing needs in seconds.
🎙️ AIPowerment Podcast dropped a new end-of-month episode featuring Märt Aro, an EdTech leader, discussing AI in education and innovation export.
🎧 Listen to AIPowerment Podcast on Spotify, Apple Podcasts, and YouTube.
Want to stay in the loop?
Straight to your inbox: practical AI updates, finance use cases, tools to try, and upcoming trainings.
I cut the noise and send only what’s actually useful.
💼 Exploring how to make AI work in finance? Let’s talk practical use cases – connect with Gerlyn Tiigemäe for expert guidance.
📚 If you would like to participate in one of my trainings or listen to speaking engagements, here are the upcoming ones:
- Training “AI võimalused finantsvaldkonnas” @ Äripäeva Akadeemia, 21.05.2025-28.11.2025 – information
- Training “AI võimalused projektijuhtimises” @ Äripäeva Akadeemia, 17.10.2025-28.11.2025 – information
- Training “AI uued rakendused”, 12.11.2025 – information
- Training “AI assistendi roll ja võimalused”, 15.10.2025 – information
