HARD MODE

March 6-8 200 participants Reimagine the future of AI x Hardware at MIT Media Lab

What if AI didn’t just control robots but it could also give humans new physical skills? At the MIT @medialab @hardmodemit hackathon, we explored using #AI + electrical muscle stimulation (#EMS) for AI to temporarily control your body, enabling people to do things they normally can't do, like playing instruments, fixing their car, and making new cocktails. 🍸 🏆 Won the Learn track 🏆

We are still feeling the energy of the past week's @hardmodemit hackathon! Special shout-out to the team behind it, all the volunteers, sponsors and of course - hackers!

One week since demo day at HARD MODE; the first Hardware × AI Hackathon at MIT Media Lab. 40+ Teams and 200+ participants building physical AI. One team gave an AI a (human) body. Others went SOFT MODE and used clay as an interface. There were plushies powered by AI, an AI that questions rather than answers, new interfaces using Pepper’s ghost, intelligent objects that live their own lives and much more… Huge thanks to our sponsors and judges from Anthropic, Qualcomm, and Akamai Technologies, whose CTO Robert Blumofe not only supported us but showed up, spoke, and sat on the jury. Grateful also to Bambu Lab, E14 Fund, Institute of Foundation Models, and all our amazing volunteers! More projects, photos, and videos coming soon. hardmode.media.mit.edu

Can tech help us fight back against the attention economy? At MIT’s HARD MODE Hackathon we built Feral Focus — an ambient AI attention companion that lives on your desk. 🏆 Won the Reflect track. Built with @MateoFerfer007 @safiware @realxdd44 Big thanks to @cyrusclarke for organizing such a hardcore event 🔥⚙️ @AHA_medialab @hardmodemit #HARDMODE

This weekend I led the organisation of HARD MODE: the first Hardware x AI hackathon at MIT Media Lab. I’ve been convinced for a while that physical AI is the big opportunity and that the most transformative applications of intelligence won’t live on a screen. They’ll be physical, embodied, sensory and in forms we haven’t seen before. HARD MODE is part of my vision of that future and a signal that the community around it already exists. 200+ engineers, hackers and researchers showed up to MIT Media Lab this weekend and spent 48 hours building towards this vision. 40+ amazing projects across 6 tracks. We had incredible support from Anthropic, Qualcomm, Akamai, Bambu Labs and more. Best of all I got to work with a great team and we managed to pull this off against all odds in 6 weeks. It was amazing to see something I had dreamt about come to life and give so much energy and inspiration to so many people. This is just the beginning. ⚙️

⚙️

cc @medialab @AHA_MediaLab Video credit: Footage by: @johary Edit by: @cyrusclarke and @johary

48h. 200+ participants. 40+ projects. One mission: Re-imagine what AI can be. This weekend we hosted HARD MODE: Hardware x AI Hackathon at MIT Media Lab. Hackers from around the globe came to build intelligent physical systems that take AI out of the screen and into the world

@hardmodemit HARD PREP 🦾

HARD at work — hackathon starts tomorrow ⚙️

HARD MODE started from a feeling: AI has collapsed the gap between imagination and physical reality, and most people haven’t noticed yet. Stuff that used to take months now takes days. I was living that. So we decided to create a space for this energy to take root. HARD MODE is a hackathon for builders who want to combine AI and physical things: hardware, embodied AI, physical systems, at a moment when making has never been more powerful. And apparently it’s the right time: 0 —> 600 applicants (closed early) 0 —> 15k followers on IG Sponsorship deals with Anthropic, Qualcomm, Akamai and a stack of others backing us. We have ~ 200 absolutely cracked builders joining us from MIT, Harvard, across the US and beyond. Craziest thing is we started working on this in Jan and built it while doing academic research. You can just make things happen. That’s the whole point of being here. We start in two days…

Robot action models shouldn't need 256 vision tokens per frame. Pi0.5 spends 400M parameters on SigLIP just to see. We replaced it with a 4.4M encoder that outputs 5 tokens — and action quality barely changes. 91x smaller. 51x fewer tokens. 7.3x faster inference.

We’re thrilled to announce Anthropic as a Platinum Sponsor of HARD MODE 2026. What happens when @claudeai meets physical things? We’re about to find out. 200 builders. 48 hours. Real hardware. March 6-8 @mitmedialab

We had nearly 600 applicants from across the world applying to attend HARD MODE - MIT Hardware x AI hackathon We selected the top 200 and currently have a 99% RSVP rate This is going to go hard⚙️