Fetching from the wire…
Public story · 2026-07-27 · high
The MIT plugin pulls captions first, falls back to Whisper transcription, then feeds Claude timestamped video frames instead of raw footage.
Why now: It added 434 of its 11.1K stars in a single day, as of July 27.
A plugin called claude-video crossed 11.1K GitHub stars on July 27, gaining 434 in a single day, according to its repository.
The MIT-licensed tool gives Claude a /watch command. Point it at a video URL or a local file, ask a question, and it answers using the footage instead of a manual transcript. For developers, it means Claude can review a recorded demo, a bug repro, or a tutorial without anyone hand-timestamping frames first.
The command behind that answer is specific about cost. It pulls native captions when they exist. When they don't, it falls back to Whisper transcription via Groq or OpenAI. Then it extracts frames with ffmpeg, dedupes near-identical ones, and hands Claude timestamped images synced to the transcript.
Four modes control how much footage Claude actually sees. Transcript-only is cheapest, an efficient mode caps around 50 frames, and a balanced mode uses scene-change detection for roughly 100 frames. An uncapped token-burner mode exists for when fidelity matters more than cost. Installing it takes one command, either /plugin marketplace add bradautomates/claude-video or npx skills add.
The four-tier cost dial is really an admission that video-to-Claude pipelines are still expensive in tokens. Uncapped token-burner mode is a bill waiting to happen for anyone who reaches for it before reading what uncapped means. Watch whether usage settles on balanced mode as the default, or whether the uncapped option becomes the plugin's biggest complaint.
Each link below shares sources, entities, or timing with this story.
bradautomates/claude-video (v0.2.0, July 1) lets agents download, frame-extract, and transcribe any video via yt-dlp, ffmpeg, and Whisper, then hand it to Claude's multimodal Read (GitHub). It ships as an Agent Skill usable across 50+ agents: Claude Code, Codex, Cursor, Gemini...
The IDE market is fragmenting, and this week drew the sharpest lines yet. Cursor 3 launched as a rebuilt agent-orchestration platform in Rust and TypeScript, replacing the VS Code fork with an Agents Window for dispatching and monitoring multiple AI coding agents. Anysphere hi...
This is the single highest-leverage spend cut I've found for daily Claude Code users, and it's one command. Run rtk init --global (or rtk init -g) and it installs a PreToolUse hook in Claude Code that transparently rewrites your Bash calls to their rtk equivalents at the proxy...
1. Use claude agents --json to build session dashboards. Claude Code v2.1.145 outputs all live agent sessions as structured JSON with status, model, elapsed time, and parent relationships. Pipe it into a tmux status bar widget or session picker script for switching between bac...
mudler/vllm.cpp mirrors vLLM's V1 / Model Runner V2 architecture in pure C++ with no Python, PyTorch, or ggml at runtime, shipping a shared/static libvllm with a stable 17-symbol C ABI, an example CLI, and an OpenAI-compatible server. Install footprint 66 MiB against vLLM's 9....
Two competing models for AI-powered security shipped on the same day. OpenAI launched Codex Security ("Aardvark") — an AI AppSec agent that builds project-specific threat models, then hunts for vulnerabilities and tests them in isolated environments. 30-day beta: 1.2M+ commits...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.