Connecting bits and atoms. Reconstructing real world events using OSINT, GEOINT and AI. Playing with spatial intelligence, world models and electric guitars.
ATX/SF
Joined December 2010
- Tweets30.4K
- Following6.1K
- Followers12K
- Likes64.6K
Pinned Tweet
Space and time reconstructed in 4D
Is this how higher dimensional beings see us - every moment at once?
Starting to understand what Vonnegut meant by humans being “great millipedes” - clearly it applies to our four legged friends too!
Geolocated 2d video → per frame 3d gaussians → each left precisely where & when it happened
Just take a minute to appreciate what 4D gaussian splats are capable of. iPhones are a perfect unit camera -- killer image quality, compute and storage on board, easy to remote control and yes GENLOCK synchronization! This is the future of reality capture.
These 4D gaussian splatting results look exceedingly crispy. Kinda wanna build a rig now. nitter.cf/mediastorm_en/status/2…
It’s pretty wild how frontier LLMs have crossed the chasm on inverse graphics.
Young founders will start decompiling reality, building RL environments, and print money doing it. Incumbents won't be too happy unless they're fast to adapt.
Case in point -- I gave AI a 3D scan of my parents old house, and it rebuilt the room in Blender. Then I turned off the gravity.
Here’s how it works:
1. ~400 photos I took in 2019 go through RealityScan and become a photogrammetry mesh
2. GPT-6 Astra rebuilds it as a clean, editable Blender scene, using the scan as reference
3. Scan textures become procedural materials. No downloaded assets
4. Every object is isolated and editable, so the physics just works
Turning messy scans into scenes modeled the way an artist would used to take days of skilled work. Huge implications for VFX, AR/VR and robotics.
Bilawal Sidhu retweeted
If you haven't seen my video with @bilawalsidhu about God's Eye View, definitely check it out.
Here's my 9th video in 14-days (trying to post 15 over 30-days to meet X's video kickstart requirements). For some reason, the backend shows I've only posted 5. 🤷♂️
I reconstructed the route Flávio Bolsonaro took to flee Brazil, 2 days after losing the presidential election and hours after a court ordered him to hand over his passports.
1. Took off from Iguazú heading toward Brazil, 6 km from the border, then turned away
2. The direct route to Miami crosses 2,416 km of Brazil
3. 2,098 recorded positions, zero inside Brazilian airspace
4. Closest pass: 69 km from Brazil, at 40,000 ft over Peru
5. 35-minute fuel stop in Guayaquil, Ecuador
6. Landed in Miami after ~7,140 km instead of 6,375
Brazil can't order a plane down if it never enters Brazil.
Context: Flávio is the son of Jair Bolsonaro, who's serving 27 years for plotting a coup. He lost the runoff to Lula on Sunday. On Monday, Justice de Moraes gave him 24 hours to surrender his passports in a money-laundering probe. Tuesday night, the family crossed into Argentina by land and got on a private jet at Iguazú.
Pixels are location data now.
I took one photo from a balcony in Austin. Then I ran it through Multiset AI against a public 3D map of the city, and it found the EXACT building and floor I was standing on.
GPS listens to satellites. VPS matches what your camera sees against a 3D model.
This gets creepy quick. Be careful posting photos out your window.
Check out @multiset_ai, and go deeper into visual positioning systems here (had a chance to work on this tech myself)
You've probably heard about GPS. But have you heard about VPS? It's the visual positioning tech that gives robots and AR glasses centimeter-level accuracy where GPS fails.
Here's how it works: youtu.be/zBHtZ1XGThA
0:00 - Intro
0:19 - How GPS Actually Works
1:28 - The Problem with GPS for AR & Robotics
2:06 - What is VPS? (Visual Positioning System)
2:38 - Use Cases: Delivery Robots & Drones
3:17 - VPS in Action: Apple Vision Pro & Meta Quest
3:53 - Scanning Your World with an iPhone (Multiset AI)
4:31 - Key Takeaways
4:48 - The Future: Spatial Intelligence
AI agents and infinite intelligence on tap make it very easy to start majoring in minor activities. When everything can get done, the hard part is discerning what's actually worth doing.
Genie 4 is out. Real-time world models make you feel like you're on set.
You can can step into a live simulation and record multiple "takes" in real-time, adjusting camera angles, lighting, sound, and action on the fly.
Here's me progressively building out a scene with the mythical black triangle the DoW shared recently as part of their UAP files.
Iconic moment in history -- stepping inside a photograph of Berlin, days after the wall opened.
Can't wait till we can actually connect our AR tracked smartphones to intuitively take control of the camera versus using keyboard controls.
Typing is fun but using voice to direct the action and the "talent" would be another great step.
Whether it's pixels only, 3d only, or a hybrid approach -- here's where I think all of this is going, and what it means for creation workflows:
nitter.cf/bilawalsidhu/status/18…
Who else is hyped about merging 3D and AI for video creation?
I'm dreaming of an AI creation tool that builds a unified 3D scene graph underpinning your story world, with diffusion applied on top like a final render pass or fancy coat of paint.
This 'dynamic' 3D scene graph would encapsulate characters, environments, and interactions. Toss it all into a large context window, and use multimodal prompting to edit intuitively and breathe life into it.
You'd be in 'director mode,' commanding your 'AI talent' like folks are doing with ChatGPT's advanced voice for improv or line delivery.
Guide your virtual 'talent', capture the best takes in a digital studio, then switch to 'editing mode' to piece it all together.
The kicker? You'd retain the flexibility to go back and make changes on a per-shot or world level.
You get all the perks of a 3D engine - cause/effect relationships, physics simulations, spatially-anchored audio - while still transforming your 'grey-boxed' world into stunning final pixels.
Eventually, you'd don an Apple Vision Pro to orchestrate the whole thing - feeling like James Cameron on the set of Avatar.
Critically, such a tool would put you in a flow state, versus chaotically jumping between an ill-tailored tapestry of tools (which is the real multiverse of madness) leaving the current set of AI video tools better suited for making memes than telling stories.
Anyone actually building this?
Same. Claude code projects is where it’s at - truly a joy. Surprised they don’t work well together. Maybe duet is trying too much to be like muse.
got me sitting at (0,0) next to bill gurley and paul graham -- thanks for the cosign grok bot
Top 100 AI accounts on X mapped.
I liked this format so I had Grok @bot make this one.
If you're unhappy with your placement on the map, blame Grok Bot, not me.
God's Eye View is becoming a time machine.
Pretty wild that we can reconstruct real events like Fleet Week, then film them in a photorealistic digital twin of San Francisco with a cinematic camera rig - all by talking to my phone.
Rendered in a browser on my macbook.
Time for another AI agent vibe check
30%OpenClaw/Hermes
15%ChatGPT Dots
15%Meta Muse
40%Grok Bot
235 votes • Final results
OSINT for tech scoops - Thariq's Anthropic exit leaked because someone checked if his corp email still existed. Tech journos about to set up a cron job to check every OpenAI email address lol
