Co-Founder & CTO at Stealth Startup | (research ⋃ prod) ⨉ (real-time ⋂ vision ⋂ graphics ⋂ AI/ML) | Ex @nvidia, @Google, @MSFTResearch | I read ⊂ the code.
Bay Area, CA
Joined March 2011
- Tweets213
- Following976
- Followers835
- Likes664
Pinned Tweet
3D reconstruction with 1734 images in 2 mins 15 seconds!
Sequential matcher with loop closure ended up dropping 4 images, but for the speedup over all-pairs matching (~10 mins) might be worth it.
Sameh Khamis retweeted
AI may well be able to write massive software projects “from scratch”, but that almost certainly isn’t the fastest / highest performance / highest security / most token efficient way to do things. Good building blocks matter.
In addition to RL’ing the agents on final task loss, we should try to get them to regularize the building blocks as well. What would the standard library look like that makes the agents most effective?
It is likely at the limit that it turns out to be some alien organization of code, but given the existing human-bias in training, I bet that in the near term it would be lovely code that old programmers would appreciate.
I suspect the same could be said for mathematics, but that is out of my lane.
Sameh Khamis retweeted
𝗘𝘃𝗲𝗿𝘆 𝘀𝘁𝗿𝗲𝗲𝘁-𝗹𝗲𝘃𝗲𝗹 𝗶𝗺𝗮𝗴𝗲𝗿𝘆 𝗽𝗿𝗼𝘃𝗶𝗱𝗲𝗿 𝗶𝗻 𝗼𝗻𝗲 𝗽𝗹𝗮𝗰𝗲: 𝗚𝗼𝗱’𝘀 𝗘𝘆𝗲 𝗩𝗶𝗲𝘄
github.com/bilawalsidhu/gods…
Sameh Khamis retweeted
Figma made design multiplayer. We made coding agents multiplayer.
Introducing Superconductor: put your team and your agents in the same session.
Run Claude, Codex, Pi, etc in the cloud, see live app previews, and ship from any device.
Give it a spin at superconductor.com
Sameh Khamis retweeted
Meet Medusa! Nine synchronized iPhone 18 Pros capture a scene from multiple angles, then four Mac Studios process the footage into a photorealistic 3D scene you can move around, reframe, relight and even slow down after the fact.
Splatting triangles too because Gaussians are so 2023!
Running here with pose fine-tuning. Follow-up papers make these opaque (w/ annealing / schedule), will def check them out.
Also implemented interaction-aware scheduling, so training runs faster when the view is idle.
Sameh Khamis retweeted
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.
Sameh Khamis retweeted
Take a minute to appreciate this demo of what an insane rig of 8 iPhone and 2 Mac Studios is capable of
Sameh Khamis retweeted
mobile-friendly 5.5GB OpenVDB, available now through your nearest @threejs provider 🙂↔️
Live: mjurczyk.github.io/openvdb/ (0.5.x)
5.5GB down to 3 streamable LOD levels:
LOD0 1.5MB (near-instant)
LOD1 8MB (mid-ground)
LOD2 50MB (as detailed as web reasonably needs)
Sameh Khamis retweeted
Kind of got carried away and made Slop New York more fun to drive around and way less sloppy 😅
mrdoob.github.io/slop-newyor…
City and vehicles are still procedurally generated at load time, and download size has gone up to 2.2MB.
Sameh Khamis retweeted
The coding interview is dead, at least for my team at Anthropic. I've seen enough to know that it no longer provides signal. I removed it from loops this month.
I have a spicy, optimistic take - I think there's an opportunity for junior engineers: felixrieseberg.com/hunger-be…
Sameh Khamis retweeted
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.
Sameh Khamis retweeted
Ghostty 1.4 is now out! Full Kitty graphics with animation (the first terminal besides Kitty to do it), scrollback compression for much lower memory usage, macOS 27 support, a new OpenGL renderer on Linux, ghostty +ssh, big performance gains, and hundreds more changes.
The highlights:
- Scrollback compression: scrollback compresses ~90+% on average, meaning your terminals use that much less live memory
- Huge memory savings across the board: no single large thing, just lots of good improvements lowers memory usage everywhere, netting anywhere from 50 to 90% less memory usage depending on workload
- Kitty graphics is now 100% complete, including animation and relative placements, after a full conformance pass against the spec and Kitty itself
- The Kitty clipboard protocol, with an image preview when a program asks to read your clipboard
- Split drag and drop on Linux, including between windows
- ghostty +ssh installs terminfo on remote hosts automatically, so things just work over SSH
- Lots of terminal emulation accuracy fixes
This release is 7 months of work from 172 contributors over 2,889 commits. It also fixes 10 security advisories, so please update.
Full release notes: ghostty.org/docs/install/rel…
Sameh Khamis retweeted
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.
Sameh Khamis retweeted
These are nice findings! Relevant on the distillation side: the original thread only discusses the issue for the soft cross-entropy loss, but the reverse KL and JSD have similar issues and fixes
Here's a summary of them 👇
Sameh Khamis retweeted
PCBs made in America. Twice as fast as Shenzhen.
Today @CircuitHub is unveiling The Grid, our factory-scale robot in Massachusetts that builds circuit boards on demand.
No changeovers, no per-design setup. Prototypes and production batches, fully automated.