What if I told you you (or your agent friends) could now super easily build your own apps, with @FlutterDev for @remarkablepaper devices.
Introducing Pluto, a remarkable way to Flutter.
How I built it with a custom device rig: /
Did this, <checks notes>, 9 years ago, before AI 😭
Made a lib that encoded JSON into a strip of RGB pixels, 3 bytes per pixel, for an external screen reader to decode, with a theoretical throughput of ~21.5 kB/s.
Fun times. Was Wow -> External only though, the font thing here
ngl having some fun with this one out on @NVIDIAAI DGX sparks - testing it still for 1 to 3 PP, but early numbers;
PP=1 single stream: 19.19 output tok/s, 129.8 ms TTFT.
PP=3 single stream: 18.66 output tok/s, 177.7 ms TTFT.
PP=3 concurrency 8: 130.25 output tok/s—1.22× PP=1.
Today we're releasing Laguna S 2.1, our most capable model to date.
It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.
Capable enough to hold its own against models many