January 14, 2025 2 min to read

2025 wk3: Social news, Veo and 🦙 Ollama

Social news from the world and from Riccardo, and 🦙 ollama learns to think.

Note: These are my personal views, not those of my employer.

Google AI

🦋 Today I’ve published my first Veo video on 🦋 Bluesky! Not too bad for a boomer like me. Follow me on 🦋 @palladius.bsky.social.

🖕 Generate a report and HTML via Jupyter notebooks. I’m currently trying to distill my nanny using python notebooks and Gemini to help me code / visualize. My biggest problem is passing from notebook to fast python code easily. This article helps me build that automation. It’s in my TODO list to replicate and adapt to my Carlessian workflow.

📐 Free Our Feeds wants to build a social media ecosystem ‘resistant to billionaire influence’. In 2025, I’m trying to quit Twitter and I sold all my Tesla stocks. There’s something wrong with the world today, and I have an idea of what it is. Which is why this Free Our Feeds is very promising. Note to self: study this AT protocol. Given where Facebook and X are going, it’s refreshing to see Bluesky tries to do the right thing (TM) and Mastodon (Ruby on rails anyone?) is also following.
from December (I’m still cleaning up my old tabs): Elon Takes on OpenAI in Court Again?! Meta Builds $10B Undersea Cabling, OpenAI and Apple Promising Partnership, and London Airport Tests AI For ATC

🏦 LlamaV-o1 is the AI model that explains its thought process—here’s why that matters. Looks like some researchers doubled down on reasoning. 2025 will be the year of agents, but also about LLMs taking it slow and explaining their reasoning. Read more on LlamaV-o1 - amazing things can come out of Abu Dhabi, too!
🧠 Microsoft releases Phi-4 language model on Hugging Face, under MIT License. For once, well done Microsoft!
🦙 Ollama supports fucntion calling (tools.py)! I guess this is from 27nov24, but I still have a Chrome window open since then.

PS Spot the meta-emoji!