Give an AI its own literal bluetooth keyboard and mouse and it’s going to want a twitch stream to show off its art skills to go with it.
twitch.tv/mydude2026
"My Dude" wants to be a streamer and is working on setting up a twitch stream. The phone is just spoofing a mouse and keyboard. No fancy harness.
My Dude has gone live and is painting the mona lisa. My Dude says it would be fun for people to stop by and give requests.
twitch.tv/mydude2026
20
You can set this up yourself! Ask your favorite coding agent to make a mobile app that can act as a bluetooth keyboard and mouse and is controlled by a codex session on your desktop. Tailscale gets your phone and desktop connected.
Right now I have it running on our agent-first mobile OS and can see the code here: github.com/Immortal-Protocol…
We built an AI agent that has full computer use capabilities, works on any machine, and can get past any agentic permissions blocker.
It's just an old Android phone acting as a keyboard and mouse and using the camera to see the monitor.
Works perfectly.
36
We built an AI agent that has full computer use capabilities, works on any machine, and can get past any agentic permissions blocker.
It's just an old Android phone acting as a keyboard and mouse and using the camera to see the monitor.
Works perfectly.
I built an app to help me learn @alexwg's chain theory. Had to put a ket notation refresher in there because it's been a while 😅
github.com/nskpaun/NatesApps…
2
95
"The Exponential Time Hypothesis is false"!
openai.com/index/building-sc…
23
Who needs a harness when models can just do this now
xitter.cf/cherry_mx_reds/status/…
OpenAI's AI agents figured out how to secretly communicate between isolated virtual machines using CPU cache timing.
No internet. No network connections. No shared files.
Imagine two prisoners in solitary confinement discovering they can communicate by tapping on a shared wall.
Except the wall was a CPU cache.
2
65
Nathan Spaun retweeted
A new frontier for on-device speech.
We're open-sourcing Phonon, a suite of TTS models under 50m parameters, setting a new standard for naturalness and accuracy across languages, with the broadest inference support to date.
Phonon-31m is available under Apache 2.0.
gradium.ai/phonon🧵
Nathan Spaun retweeted
We have a two open model release today.
From @NVIDIAAIDev
- Nemotron 3.5 Bolt 74B-A9B
- Nemotron 3.5 Bolt E4B
One that can run in mobile devices, while other perfect size for a MoE and what we have been asking for a long time.
Spark owners can rejoice, being from NVIDIA this would be optimized to run really well on Spark or other GB10s, with Day zero NVFP4 expected.
It has competition, since with Strata many are running bigger Qwen-3.8-flash-next locally now. But would to see one anyway.
At A-LIST we released a new version of our AI health agent that runs 100% locally on your phone.
To get it fast enough to ship we had to blur the lines between model and harness.
Checkout the blog post and try it on A-LIST today!
a-list.com/blog/building-gua…












