AI, Graphics, and Doodling.
I am a Senior Staff Research Engineer at Meta Reality Labs, working on real-time graphics and visual computing. This site keeps growing the same three things: models run and probed on machines you can point at, renderers written to find out how light actually reaches a pixel, and a crayon-doodle shelf of bedtime picture books drawn to a four-year-old's specifications.
Latest writing
All 30 pieces →-
Hold a Key and Talk to Jarvis
One night after Cadabra 1.2, my dictation app learned to lean toward my own words, to edit text by voice, and to type as I talk. Then it got a third key that hands what I say to an agent. My home voice assistant, Jarvis, is now one press away on the Mac, through a plugin that does everything but listen, and that window soon had tables, pictures and follow-ups.
Nov 19, 2026
-
Abracadabra, It Learns My Words
MyWisper 1.2 learns from the words I fix after dictating, but only the fixes that sound like what it heard. A week of trying to make the speech model itself adapt to my voice showed what that takes. Then it got a new name: Cadabra. Still nothing I say leaves the Mac.
Nov 18, 2026
-
My Dictation App Now Fixes What It Mishears
Two days after I released MyWisper, a quick test of Whisper turned into something better: a language model inside the app that reads what the speech model heard and fixes names and misheard words from context, in about a third of a second. Like before, nothing I say ever leaves the Mac.
Nov 12, 2026
-
Teaching an Old Ender 3 New Speeds
My Ender 3 V2 took almost three hours for a part the new printer makes in eighty minutes. One evening I moved it from its stock firmware to Klipper: a fresh OctoPi first, firmware built in Docker on the Mac, PID and extrusion checks, and an accelerometer on the Pi to measure how the frame shakes. The printer can now accelerate almost six times as hard.
Nov 12, 2026
-
Dictation That Stays on My Mac
Hold a key, talk, let go, and the words appear wherever the cursor is. My first version, a Python app Gemini wrote, didn't quite work: its menu bar icon never appeared, and any web page could read what I'd dictated. So it became a native Mac app, with a small speech model that transcribes five seconds of speech in 50 ms and never touches the network. It's now free on GitHub.
Nov 10, 2026
Selected work
All projects →-
AOS: Agent Operating System
A research prototype for agent collaboration through runtime primitives: artifacts, messaging, capabilities, and scheduling.
-
Jarvis Voice Assistant
An Android voice assistant with local Qwen inference, speech models, home-control tools, and a background update pipeline.
-
Cadabra
A Mac dictation app: hold a key, talk, and the text lands at the cursor. Speech recognition and an optional language-model correction both run on the Mac, and it learns the words I fix.
-
Generative Facial Keypoints
Exploring image-conditioned diffusion and flow matching for facial landmarks with frozen DINO features and transformer coordinate models.
-
Wukong & Friends
Crayon-doodle bedtime picture books for young readers, with free reading and two-up booklet PDFs, published from a private repository against an Ed25519-signed manifest.
-
Bread
Six loaves, from a basic white bread to rye: recipes with before-and-after photos, in the order they were baked.
Also runs in your browser: Ray tracer · ImageProcJS · Fractal explorer · MineSweeper · Multilinear Reconstruction