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: AI agents are starting to do real work
This week’s biggest theme is clear – AI is moving beyond chat. It is starting to handle real tasks, use tools, and work more independently. We saw this in several updates, especially from Perplexity and Anthropic. Instead of only answering questions, these systems are starting to help with budgeting, coding, planning, and task management.
A good example is Perplexity Computer, which can now connect to financial data through Plaid. That means users can ask for things like a budget view, a loan repayment plan, or a net worth tracker in plain language.
At the same time, Anthropic’s Managed Agents shows the same direction from another side – helping companies build agents that can work for hours and share tasks with each other.
Why does this matter? Because this is where AI starts to feel less like a smart assistant and more like a digital worker. For everyday users, that could mean less manual work. For businesses, it could mean faster workflows and new ways to automate routine tasks. But it also raises bigger questions around trust, safety, costs, and how much access we are ready to give these systems.
In simple terms: the AI story is changing. The main question is no longer only “Which model is smartest?” Now it is also “Which AI can actually get useful work done?” That shift may define the next stage of AI adoption.
LLMs & AI Models
- Perplexity added Plaid integration to its agent, enabling financial data access and automated budgeting, tracking, and planning tools via prompts.
- OpenAI launched a $100 Pro plan with 5x more Codex usage, targeting developers building agent-based coding workflows.
- ChatGPT is now available in Apple CarPlay, enabling hands-free voice interaction while driving.
- Meta introduced Muse Spark, a multimodal reasoning model handling text, voice, and images with competitive benchmark performance.
- Anthropic released Claude Managed Agents, simplifying how developers build and run long-running, collaborative AI agents.
- xAI is training 7 new models on Colossus 2, including massive 6T and 10T parameter systems.
- Z AI launched GLM-5.1, an open-source coding model beating GPT-5.4 in benchmarks and running 8-hour autonomous tasks.
- Anthropic secured 3.5 GW compute capacity and scaled revenue to $30B, showing rapid enterprise AI demand growth.
- Anthropic restricted third-party agent tools in Claude, shifting to usage-based pricing and sparking community backlash.
New Tools
- HeyGen launched Avatar V, creating realistic video avatars from a 15-second clip while reducing identity drift and improving lip-sync accuracy.
- Google released AI Edge Eloquent, an on-device dictation app that converts speech to text in real time without cloud usage.
- Netflix introduced VOID, a physics-aware video editing framework that removes objects while adjusting the scene realistically.
- Pika Labs launched PikaStream 1.0, enabling AI avatars to join video calls with real-time voice cloning and interaction.
👉 Explore these tools: HeyGen Avatar V | AI Edge Eloquent | Netflix VOID | PikaStream 1.0
Other Quick Picks
- Amazon is investing $200B in AI, with AWS AI reaching $15B revenue and strong demand for chips like Trainium and Graviton.
- Oxford researchers built an AI that predicts heart failure 5 years early with 86% accuracy using CT scans.
- Canva acquired Simtheory and Ortto to expand into AI agents and marketing automation workflows.
- Anthropic launched Glasswing, a cybersecurity coalition using AI to detect critical vulnerabilities across systems. I wrote about it too, because this is super scary.
- OpenAI, Google, Anthropic are collaborating to limit Chinese AI model distillation via shared security efforts.
- OpenAI proposed a new social contract, including AI taxation, public funds, and a possible 4-day workweek.
- The New Yorker revealed new details about Sam Altman’s leadership issues, based on internal documents and interviews.
- Iran threatened a US AI data center in Abu Dhabi, highlighting growing geopolitical risks around AI infrastructure.
- Anthropic acquired Coefficient Bio to expand into AI-driven drug development and life sciences.
🇪🇪 AI News from Estonia
- Eesti.ai launched 15 AI projects, aiming to train 100,000 people and bring AI to education, business, public sector, and healthcare.
- A KPMG study shows 69% of large Estonian companies limit AI use to few employees, despite clear productivity gains.
- Estonia blocked 5.6M scam calls in Q1 2026, as AI-driven fraud becomes more massive and complex.
🎙️ AIPowerment Podcast dropped a new episode of AI news covering OpenAI’s strategy shift, Claude’s growth, AI tools, and Estonia’s AI developments.
🎧 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, next trainings for beginners @ 14.05 – information; for intermediate level @ 13.04 – information
- Training “AI assistendi roll ja võimalused” by areng . ee @ 16.04.2026 – information
- Training “AI assistendi roll ja võimalused” by areng . ee @ 16.04.2026 – information
