On‑Device AI Builder

Building offline AI. Private by default.

[On‑Device AI Builder @raffa85786]

@boardyai 🤣

[On‑Device AI Builder @raffa85786]

@elonmusk @enzo_gte It must be interesting if Elon said it 🤔

[On‑Device AI Builder @raffa85786]

@DoanHa417407 @elonmusk @enzo_gte 😆

[On‑Device AI Builder @raffa85786]

@sama Will try it.

[On‑Device AI Builder @raffa85786]

@MrBeast 185,000

[On‑Device AI Builder @raffa85786]

@GlennSonna I need it to release model memory between runs on 4GB devices.
Thanks.

[On‑Device AI Builder @raffa85786]

@GlennSonna versions don’t match, so I need to confirm one thing:Is XybridModel::unload() actually callable from Dart in the current xybrid_flutter release (0.2.0‑alpha)?If yes, what is the Dart method name?If no, can unload be exposed in the next binding update?

[On‑Device AI Builder @raffa85786]

@GlennSonna But in the Flutter binding:I don’t see any Dart‑side method that maps to unload()The generated BoltFFI surface doesn’t appear to expose itAnd in the 0.2.0‑alpha diff, you mention that Result<(), E> exports (warmup/unload) can be mis‑generated as void if the boltffi CLI/runtime

[On‑Device AI Builder @raffa85786]

@GlennSonna Hi Glenn quick binding‑surface question.I can see in the Rust SDK that:rustCopyXybridModel::unload()is a real method that resets the executor and drops the ONNX/GGUF sessions (so it genuinely frees model memory).

[On‑Device AI Builder @raffa85786]

@GlennSonna Does the Xybrid runtime handle memory pressure automatically on 4GB Android devices when multiple BYOM models are loaded, or does the app need to manually release models? I can see there's no unload method exposed in the flutter binding.

[On‑Device AI Builder @raffa85786]

@JakeFleshner Sovereign AI

[On‑Device AI Builder @raffa85786]

@ShadowVeight Nice one, Adam, wishing you the best with https://t.co/TqSPnPxF7R and the pre‑seed push.

[On‑Device AI Builder @raffa85786]

@Exposes_ Respect, Nick. You’re right. The hardest part was pushing through without knowing if I’d even put things together right in the first place. Still shipping, still learning, still moving forward.

[On‑Device AI Builder @raffa85786]

@markjivko Right now, the biggest hurdle is making sure every part of the system fits together cleanly. When you build something complex without a background, the hardest part is not knowing if you’ve assembled it right until much later. I’m past the worst of that now, just refining now.

[On‑Device AI Builder @raffa85786]

@elonmusk

[On‑Device AI Builder @raffa85786]

@calvinluvsrice @1Umairshaikh Will get back to you on this soon.