Jason Hadjioannou (@Jasonio) header
Jason Hadjioannou (@Jasonio) profile picture

AI Scientist (pre-OAI) *building stuff

London 🇬🇧, 🇬🇷
Joined October 2015
Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)Photo by Jason Hadjioannou (@Jasonio)
Jason Hadjioannou retweeted
I'm really excited to announce my own store for Nintendo 3DS. You can keep my ports up to date, download other ports from the community, and much more right from your console. Git: github.com/EstebanPdN/forge.… Discord: discord.gg/GSkDRMGMy
53
487
11
5,598
109,900
Jason Hadjioannou retweeted
🐾 AtomDog — A Tiny Open-Source Robot with Big Personality! 🤖 Created by se kino, this 3D-printed quadruped pairs M5Stack #AtomLite with the #AtomicMotion Base for browser-based calibration and control—no app needed! 📱✨ Walk, sit, and wave a paw with four micro servos 🐕 Fine-tune servo calibration and gait settings in real time 🎛 Switch between English, Japanese, and Chinese in the WebUI 🌐 Build and customize with open-source firmware and 3D models 🛠 👉 Explore the project on Hackster.io: bit.ly/4iVLA1g #M5Stack #AtomDog #ESP32 #Robotics #OpenSource #3DPrinting #MakerProjects
7
73
11
455
17,137
Jason Hadjioannou retweeted
Quick one on Living Weights for anyone new here 👇 Every open model you download is frozen. The chat ends and it forgets you. Living Weights changes that. Teach it something, it saves it into the weights, and it still knows it after a restart, even on another machine. No KV stashing, no MD files. It learns between replies, on your own hardware. Still experimental, and @petruspennanen and I are still pushing it. What would you teach your model first?
46
14
2
252
9,471
Jason Hadjioannou retweeted
106
572
35
4,838
95,288
Jason Hadjioannou retweeted
How are LLMs built? I sat down with @jeremyberman, an Anthropic researcher, for a nearly hour-long masterclass on how it all works. Extremely proud of this video, the feedback has been tremendous. Thank you so much for doing this with me, Jeremy.
25
39
7
453
102,254
Jason Hadjioannou retweeted
We just killed restaurants. Introducing microwave.io Everyone can now cook their own food at home in just 60 seconds.
56
36
9
885
44,546
Jason Hadjioannou retweeted
Introducing Omarchy Pocket. Real Omarchy Linux on a $99 dual-screen handheld. We rebuilt the Wayland GUI with PocketJS Applets, made it fluid on hardware that can’t even run Android, and rethought how Linux should work across two screens.
25
73
25
847
46,548
Jason Hadjioannou retweeted
Either get into wearable AI, hardware, or farming...
58
26
7
646
29,409
Jason Hadjioannou retweeted
We should start a Homebrew Robotics Club in Palo Alto.
15
4
2
124
5,757
Jason Hadjioannou retweeted
Unbelievable from the community I’m blown away ❤️
Almost there, go @ashxhart, go ✌🏻
24
6
2
221
6,050
Jason Hadjioannou retweeted
TensorFold for Mac 1.0.0 is out, which is our first stable Mac app release, with Linux and Windows to follow shortly. What's changed: • A compact Ask TensorFold bar, with window selection, a preview of what you're sharing and local text extraction where available. • Close the main window and keep TensorFold running in the background, with the mesh icon in your menu bar. • Connection health and inbox improvements, so you can see account status and provider errors. • Portuguese interface support. • Engine updates without replacing the whole app, with the option to return to the bundled engine. The app includes engine 1.2.0. You can choose to install 1.2.1 through Models → Details → Engine updates. Activation waits until local work has finished. Thank you to everyone who's tested the betas, reported bugs and supported the project. tensorfold.dev
6
8
2
132
6,743
Jason Hadjioannou retweeted
Wild idea. The UK does not have any AI labs creating models. Who would be up for banding together in the UK to see if we can create a open community model 🤷‍♂️ Maybe it will be trash but I for one would love to learn and try. Dm me if you are game 💪🏻
196
22
4
577
47,860
Jason Hadjioannou retweeted
you are the moat
3
3
1
55
1,999
Jason Hadjioannou retweeted
if we launch enough of these (for <$100 each) we'll literary have our own grid. and then can text all the homies via balloon
We've designed a custom stratospheric balloon platform weighing just under 10 g. anyone can build and launch one by following the docs stratolink.org
41
88
8
1,534
51,259
Probably the most valuable plastic parts in the worlds
8
5
1
77
8,761
Jason Hadjioannou retweeted
You can build your own humanoid robot.
Asimov 1 is an open-source reference humanoid robot. It ships as a complete DIY kit. $20,000 total, delivered to your door. menlo.ai/asimov-1
92
247
51
2,496
115,706
Jason Hadjioannou retweeted
This is by far one of the coolest maps of any game I've ever played I know this place like it's my home
224
492
93
7,225
2,088,233
Jason Hadjioannou retweeted
Consciousness is the original vibe coding algorithm
63
47
8
523
16,790
Jason Hadjioannou retweeted
🚀 Excited to introduce StreamRender-H3 — bringing real-time interactive worlds to life with diffusion models! Built on Sol-Engine, we post-trained @MiniMax_AI H3 Ref2VA into a real-time streaming neural renderer. 🎮 Code drives the world. A game engine handles physics, object states, and player interactions, while H3 generates the visuals. ⚡ Real-time streaming: 2 denoising steps, 24 FPS, and ~273 ms per chunk on 8×GB200 GPUs. 🎨 One world, endless appearances: Change the reference image and prompt to render the same interactive world in entirely different styles. 🔥 Powered by streaming post-training, distillation, a fast causal H3 VAE, and full-stack inference acceleration. Check out our demo and code! 👇 demo: nvlabs.github.io/Sana/Sol-En… code: github.com/NVlabs/Sana/tree/…
37
118
32
904
72,726
Jason Hadjioannou retweeted
Researchers at MIT developed Draft, a tool that lets engineers design and customize robots by adjusting parameters such as limb length, body size, and motor power. It uses real-world hardware data and reinforcement learning to estimate robot feasibility and test performance in simulation. The tool helps researchers explore robot designs faster without rebuilding CAD models for every change.
7
78
8
617
29,662