

레밀리아 한국 지부에 오신 것을 환영합니다. ☆ (주) CULT : cult.inc ☆ ♡ 흑색작전부서 : t.me/remilia_korea ♡ WE ARE EVERYWHERE. WE ARE EVERYTHING. 레밀리아 월드 오더
͏͏͏͏͏͏
28.5K posts

@Chris_In_Shell
♡ Remilia One Thousand Forever ♡ @remiliacorp333 ♡ ☆ @MiladyCult ♡ $CULT ♡ 🇰🇷 @miladykorea BLACK OPS DIVISION ☆ Rep. of @Arkenstone_Labs


레밀리아 한국 지부에 오신 것을 환영합니다. ☆ (주) CULT : cult.inc ☆ ♡ 흑색작전부서 : t.me/remilia_korea ♡ WE ARE EVERYWHERE. WE ARE EVERYTHING. 레밀리아 월드 오더

the first octra program with fully public inference for SmolLM2-135M (training, weight loading, and state are fully public) currently, it's only for informational purposes, because the full cycle is expensive (4k OCT) since it performs about 1 billion FP64 arithmetic ops so, wait for an update of the webcli with an interface for interaction via a wrapper (will be available tomorrow) now you can look at the process of loading weights: octrascan.io/epoch.html?id=… (using this epoch as an example) example exec with resp: octrascan.io/tx.html?hash=a… * the program is verified and completely open





"You were early, but now everyone is building privacy, decentralized compute and verifiable vibecoding, so how will you differentiate?" Well, announcing isn't doing. Doing takes years, and is much harder with tons of legacy than starting from scratch 5 years ago, actually.





