a quick sketch of prompt composition for image generation models (in this case, the wonderful Midjourney!). individual prompt tokens are weighted based on distance to the creation button.

looks like visionOS 2 has enterprise APIs for specifying object tracking detection rate and for increasing performance headroom, along with a new API for requesting hand anchor data at a given timestamp (developer.apple.com/documentation/…). hopefully this leads to >30hz hand tracking data!

up in maine working on computational jewellery for the next two weeks

having a lot of fun running around in sdxl-turbo's encoder_hidden_states to steer visual style.

🛝 I want to invite you to come play with us! Today we launched our Multi Modal Playground on Cloudflare Developers 🧡

love this work so much! feels like the next step on from apple’s fantastic widget & complication work — data candy, sticker apps, chewable compute. stick them to your carplay dashboard, leave them scattered around your visionOS space like marbles.

Cascading image variations

🪸 image variation UI 🪸 pull out descendants, evolution via artful selection

sci-fi so close you can almost touch it

spoke with Tyler Angert about his ideas here a couple of months back — can’t wait to see them come to life! very excited to play with true creative tools for software development.

more computational silversmithing. 3d diffusion growth finger cap and parametric sine ring cast in 925 sterling.

exploring aizawa strange attractors with ble midi and a mems laser projector

exploring aizawa attractor parameter space through gestural input