had an amazing time at the bracket bot hackathon this weeekend. we built a tldraw style no-code interface to program the robot to move, think, and talk. shoutout my team (Arjun Krishna, Amitav, and Joe) and brian-machado-high-inference for making this possible :)

this weekend i flew to waterloo to build my first robot. we built police bot—a single-purpose surveillance bot. i added voice commands for the drive system, set up leds to signal forward/back, stop, and left/right, and integrated a vision mode for emergencies where the bot

Over last weekend, my friends & I worked on building a bracket bot from scratch. We used 3D printed grippers with flexures to grip a box, an arm that extrude to press an elevator button. To recognize images of the elevator, Roboflow helps trains the bot to recognize the elevator.

Building a robot from scratch in a weekend is now possible. Robotics + AI + builders are going to change everything about the world around us.

More Bracket Bot content! From our Oculus VR-to-bot hack, from last weekend’s hackathon, ft. a clip of our bot moving via instructions from the Oculus headset. We also got some speech-to-text translation working just at the end! (Though hit some bugs during demo time)

Saeejith Nair you can just parse things

annotations should feel more like “curating a garden” vs transactional labeling

MASt3R-SLAM is so good, it just works

HEY YOU 🗣️🗣️ brian-machado-high-inference built a messenger from the future 🤖

> spawn in toronto > grind in waterloo > explore startup bubble in sf > explore tech/culture bubbles in nyc > rediscover love for toronto as a healthier and more scalable version of nyc/sf > endgame in toronto

Make beautiful (annotation) software

You're in the arena, Socratica is building it

In 20 years, the idea of the Great Stagnation will be a fantasy of a bygone era. People will be shocked that at some point in the past we worried about growth and dynamism. They will laugh at the fact we thought progress had stalled. "Look around!" they will say. "Look at at

If you think Brampton is good, wait till they drop Scarborough

Super excited to have Akira Yoshiyama ⁂ give a talk on LADDER this Thursday at 4pm ET! If you're on campus, join us in EC4 2101A or tune in online: calendar.app.google/EBy6zMG4g21tgb…

o3 seems to slightly mess up numbers a lot more than o1 (e.g. 58.6->56.8, 63.8->63.4). amazing model otherwise

I find it amusing these posts were published a day apart. The apparent contradiction pivots around how easily the task context can be factorized. Devin finds coding tasks tend to be less factorizable than research tasks, and therefore advocates against the multi-agent approach