🧠 Robots, Crime Busters & The AI Tools You Actually Need

A New York-based startup called Mecka AI just raised $60 million to do something that sounds almost obvious once you hear it: teach robots how to move by recording how people move. Using body sensors and iPhones, Mecka collects human motion data — hand gestures, walking patterns, physical interactions — and uses it to train what the industry calls "embodied AI," the kind that needs to exist in a physical body to do physical things. Founder Josh Gao says he's convinced useful robots can be deployed today, not in some distant future.

The bigger story here is data. Every robotics company is racing to get the training data that will make their models actually work in the real world — and Mecka is betting that human motion, captured at scale, is the missing piece. It's an interesting bet, and the investors seem to agree.

Pax, an AI-native public safety company based in São Paulo, just closed a $40 million seed round co-led by Greenoaks and Benchmark — one of the largest seed rounds in Latin American history. The numbers behind the raise are worth pausing on: in its first large-scale deployment across 30+ Brazilian cities, Pax's platform cut violent crime by 27%, doubled police efficiency, and improved public perception of safety by 59%, all in six months.

What Pax actually does is connect cameras, police records, and criminal databases into a unified real-time intelligence platform. Officers get investigative leads and alerts; the AI handles the data grunt work — cross-checking footage, matching plates, surfacing patterns across disconnected systems. More than 2,000 criminal cases have been resolved through the platform so far.

The founder, David Peixoto, frames it clearly: the tool researches, the officer decides. That framing matters, especially in a conversation where AI and law enforcement together tend to raise questions. The platform logs every single query to a named, authorized operator, which is the kind of transparency that makes a tool like this defensible.

ChatGPT rolled out a full-screen editing mode for longer pieces, so you're no longer writing a 3,000-word draft inside a chat bubble. Longer documents now get the room they need. The update also adds a Library feature, so you can save your work and come back to it later — a small thing that makes ChatGPT feel more like a writing environment and less like a conversation you'll lose when you close the tab.

If you write longer form content using ChatGPT, this is worth checking out immediately.

Rodeo is an AI-powered video editing assistant built for anyone who works with large amounts of footage. You upload up to 50 hours of raw video, and Rodeo processes and understands all of it — so you can search in plain language across your entire library, organize clips by people, actions, or emotions, and assemble rough cuts without manually scrubbing through hours of footage. From there you can build storyboards, swap clips, and export to your usual editing software via EDL or OTIO.

HyperFrames is an open-source framework that lets you create videos by writing HTML, CSS, and JavaScript — or by just describing what you want to an AI coding agent. Think of it as treating video like a web page: you define your scenes, animations, and layout, preview it in a browser, then render it as an MP4. AI agents like Claude Code, Cursor, and Gemini CLI can write and edit the whole thing from a plain language prompt.

This one is geared toward developers and technically-minded creators who want to build repeatable, scalable video workflows without touching a timeline editor. Open source, Apache 2.0 licensed.

Most of us have lost the manual, forgotten what we did with it, or never opened it in the first place. Here's a smarter approach using Gemini — one setup, permanent use.

A few alternatives worth knowing: NotebookLM works really well for this too — create a dedicated notebook and drop all your manuals in. Some people also just type in the model number and ask AI to find the guide, which works fine for a one-off fix. But if this is something you'll need repeatedly, having a persistent Gem or notebook saves you the setup every single time.