Research & Blog
Research and notes
Papers, benchmark results and engineering notes from the team.
Featured
-
VideoDB leads a 100-hour visual moment retrieval benchmark
VideoDB against AWS Nova, Twelve Labs and five open embedding models, on an open dataset with open queries.
Read blog -
How a robot remembers
How a robot keeps an index of everything it has seen, and pulls back the few seconds a task needs.
Read blog -
Search over the Visual World
Persistent visual memory, layered indexes and source-grounded evidence for agents over cameras, screens and archives.
Read paper
All posts
-
Can Claude and ChatGPT Watch Videos? Yes, Here's How Read blog
-
How to add video to an app you built with Cursor Read blog
-
How to turn a video into a link with Claude or ChatGPT Read blog
-
Build your own video social-listening tool with VideoDB Read blog
-
A Big Four accounting firm sets up its global capability centers with agents Read blog
-
100,000 hours in under a week for one of the largest video data providers Read blog
-
How a robot remembers Read blog
-
How to evaluate a video AI platform: a PoC checklist Read blog
-
Understanding robotics datasets: inside LeRobot, RLDS and MCAP Read blog
-
Frame access that keeps research moving Read blog
-
What is Physical AI? AI that acts in the real world Read blog
-
VideoDB leads a 100-hour visual moment retrieval benchmark against AWS Nova and Twelve Labs Read blog
-
Agentic video analysis: what works and what doesn’t Read blog
-
Why a search tool is not enough for an LLM working with video Read blog
-
A query engine for robot video Read blog
-
Can a VLM annotate a robot run? Results on a revised WGO-Bench Read blog
-
Giving an agent a YouTube video it can search Read blog
-
Claude edited our launch video with VideoDB skills Read blog
-
Twelve Labs alternatives, compared Read blog
-
Video RAG: how it works Read blog
-
Agentic Streams: an agent researches a topic and streams you the briefing Read blog
-
Search over the Visual World: persistent visual memory, layered indexes, and source-grounded evidence Read paper
-
Real-time visual perception for agents Read blog
-
From an RTSP stream to events an agent can act on Read blog
-
Unlock the Footage: July 2026 project showcase Read blog
-
What video retrieval benchmarks taught us about ground truth Read blog
-
DeepSeek’s visual primitives and the missing layer Read blog
-
JEPA, from language models to world models Read blog
-
TinyFish × VideoDB: the web your agent browses, as video it can search Read blog
-
Agent Forge: June 2026 project showcase Read blog
-
Episodic memory for agents Read blog
-
Video was built for playback, not perception Read blog
-
MP4 is the wrong primitive Read blog
-
Perception is the missing layer Read blog
-
Bloom: record your screen, search it later Read blog
-
Focusd: a coach for your workday Read blog
-
A camera feed for your OpenClaw agent Read blog
-
Bloom, AI-ready screen recorder An agentic Loom alternative where recordings become inputs for AI. View app
-
Focusd, productivity tracker Screen memory for understanding where your workday went. View app
-
OpenClaw Monitoring CCTV for computer-use agents. View app
-
Infrastructure that sees and edits Read blog
-
Why agents are blind today Read blog
-
A strong general-purpose VLM still fails on a chessboard Read blog
-
Deep Search: finding exact moments in video Read blog
-
Pair Programmer: live screen and audio context for coding agents Read blog
-
Global Hackathon: May 2026 project showcase Read blog
-
Evaluating video VLMs on your own task Read blog
-
Dispatch #001: playable evidence, thinking tokens, agentic streams Read blog
-
How I built Call.md on VideoDB Read blog
-
Call.md, meeting copilot Turn meetings into live agent loops. View app
-
Cool QR Codes Studio A no-login QR code studio with scanability checks. View app
-
Dictate Local-first macOS dictation that pastes speech into any app. View app
-
Reuse HTTP connections in Python services Read blog
-
Caching CORS preflights: hundreds of milliseconds per call Read blog
-
Do thought streams matter? Evaluating reasoning in Gemini vision-language models for video scene understanding Read on arXiv
-
A Postgres backup is not real until you restore it Read blog
-
The 6 MB Lambda limit: compress before, not after Read blog
-
Searching the slides in a conference recording Read blog
-
NFL game analysis: 80% fewer VLM hallucinations Read blog
-
Smaller Python ML containers Read blog
-
VideoDB × TwelveLabs: search across your video library Read blog
-
VideoDB × LlamaIndex: video in your RAG pipeline Read blog
-
Benchmarking vision-language models on optical character recognition in dynamic video environments Read on arXiv
Nothing here yet.


