Sam

Want the full work? Paper: arxiv.org/abs/2604.11177 Code: github.com/video-db/gemin…

Our paper "Do Thought Streams Matter?" got featured in a YouTube video 👀 Quick watch, simple breakdown, and some cool insights on Gemini video reasoning. Watch here: youtube.com/watch?v=8HN19n…

@alexabelonix Thanks! 🙌 glad it resonated

@shar_dev7 love this approach.

Can video AI really reason about what it sees, or does it just sound confident? We explored this in our new paper on Gemini vision-language models for video scene understanding. Some interesting results came out of it 👀

If you work on VLMs, video understanding, or multimodal reasoning, this may interest you. Paper link: arxiv.org/abs/2604.11177 Github repo : github.com/video-db/gemin…

This matters because a lot of people assume more reasoning = better answers. But for video models, the story may be more nuanced. Our paper looks at this closely and helps make sense of how reasoning works in real video understanding settings.

@ashu_trv Interesting perspective on reasoning efficiency in VLMs.

We just released a new benchmark looking inside the "black box" of Gemini 2.5 reasoning for video understanding. Does "thinking more" always lead to better results? The answer is more nuanced than you’d think 💭

It's happening. New leak from Anthropic appears they are building a full-stack app builder inside Claude. They are coming for everything.

If you wanna try this open source tool 👇 github.com/video-db/pair-… #AI #Agents #DevTools

No logs No screenshots No guessing Just context This is what happens when agents can see + hear + remember Feels like debugging with memory

Tried something new today 👀 My auth flow broke Instead of checking logs I asked: “what was I doing before this?”

Try it 👇 github.com/video-db/agent…

Videos → searchable Videos → interactive Videos → usable This is how video should work

What if your video wasn’t just something you watch… But something you can talk to That’s Agentic Videos 👇

Now featured on Agent Community Making agents understand video, screens, and real-world context Check it out 👇 agentcommunity.org/m/videodb

AI agents can read text But what if they could see and hear too? That’s where VideoDB comes in Building the perception layer for agents 👇