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: Meta buys Manus for $2B: ambition, agents and a price worth paying?
What started as a WeChat side project by Chinese student Xiao Hong has now become one of the boldest acquisitions in the AI world: Meta is acquiring Manus, a general-purpose AI agent startup, for over $2 billion. At just 33, Xiao – nicknamed “Red” – has built an autonomous agent platform powerful enough to catch Mark Zuckerberg’s attention and strong enough to ignite investor FOMO across continents.
Manus isn’t just another chatbot. Launched in early 2025, the platform stunned observers by completing tasks like coding, research, and data analysis without human hand-holding. Millions of users later, and with 147 trillion tokens served, it’s a prime example of how agent-based AI is stepping into the mainstream. Meta says the team and product will now integrate across its ecosystem, bringing autonomous workflows into Meta AI and enterprise tools.
But with a price tag that exceeds what most startups can dream of, some are asking – was this overvalued, or is it a bet Meta can’t afford not to make? After all, AI agents are becoming the next frontier – an area where Meta lags behind Microsoft’s Copilot and OpenAI’s ecosystem. With Manus, Meta gets not just a product, but a fast-moving team, proven demand, and a founder who knows how to survive in competitive, regulatory-heavy markets.
At the heart of this acquisition is a clash of visions: China-born innovation meets Silicon Valley scale. It’s also a strong signal that the AI race is not just about models anymore – it’s about what those models can actually do. If agents really are the next AI layer, Manus might just be the best early bet Meta could have made – even if they paid a premium for it.
LLMs & AI Models
- DeepSeek introduced a new method (mHC) to boost training stability and logic performance in large models with minimal compute overhead.
- OpenAI is merging voice teams ahead of launching a Jony Ive-led AI device, enabling fluid, interruption-resistant voice conversations.
- LMArena community voted Gemini 3 Pro as best in text, vision, and search, while Claude Opus and GPT-4 took top spots in other domains.
- GPT-5.2 Pro hit 29% on Tier 4 FrontierMath, beating Gemini 3 Pro by 10% and setting a new benchmark for AI scientific reasoning.
- Claude AI lost money in a Wall Street Journal experiment where it was tricked into giving away free PS5s, showing vulnerability to social manipulation.
- Meta unveiled a self-improving AI that finds and fixes its own code errors without human data, outperforming human-trained baselines.
- Claude Opus 4.5 handled uninterrupted 5-hour workflows, making it the most durable model in long-form task performance tests.
- xAI released Grok Business and Enterprise, with admin tools like SSO, SCIM, and custom encryption layers for secure team deployment.
New Tools
- Adobe partnered with Runway to integrate Gen-4.5 video models into Firefly, expanding Adobe’s capabilities in AI-powered video editing.
- Notion revealed an AI-first workspace with tabs for chats, pages, and inbox, plus admin controls and a credit-based usage system.
👉 Explore these tools: Adobe Creative Cloud | Runway | Notion
Other Quick Picks
- Instagram’s head says users now prefer private messaging over public posts, raising challenges in identifying AI-generated media. I create a post and poll about it here.
- Claude AI kept a tomato plant alivefor over a month without human input, managing a physical environment through automation.
- OpenAI employees earned a record $1.5M in stock compensationin 2025, marking the highest equity payout in tech startup history.
- SoftBank completed a $40B investment into OpenAI, betting on a future IPOby selling off Nvidia and T-Mobile shares.
- Satya Nadella said AI is moving from hype to measurable impact, emphasizing real-world integration over flashy breakthroughs.
- 21% of YouTube recommendations to new users are low-quality AI-generated slop, designed to boost views and ad revenue. Read the whole report here.
- Andrej Karpathy admitted feeling left behind as a programmer, highlighting how fast AI tools are reshaping software development.
- OpenAI is hiring a Head of Preparedness for $555K to assess risks, build safeguards, and address growing AI safety concerns.
- Retailers tested AI to fight return fraud during holidays, flagging suspicious activity before issuing refunds in the $76.5B scam sector.
🇪🇪 AI News from Estonia
- Wise and RTK HR leaders are investing in internal AI training and automation to support routine tasks while emphasising data security and critical thinking. Listen here.
- TalTech professor Raivo Sell says autonomous vehicles in Estonia may launch within 5 years, but only in restricted zones like robotaxi services.
- Estonia plans to align X-tee with Gaia-X, positioning the country to lead in public sector data space development across Europe. Read here.
🎙️ AIPowerment Podcast is on a small winter break but you are welcome to listen the latest episode by Gerlyn Tiigemäe and Sandra Reivik, reviewing 2025’s top AI trends and lessons for 2026.
🎧 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 in 2026 – information
- Training “AI võimalused projektijuhtimises” @ Äripäeva Akadeemia, next trainings in 2026 – information
- Training “AI assistendi roll ja võimalused”, 03.02.2026 – information
