BubuAi

What happens if you use 5.6 Luna light fast mode?

@lightsilver323 inkmarshal.com

drop your startup link get some traffic

@lightsilver323 lunerylab.com

Good news: Opus 5 is out, and it seems extraordinary. Bad news:

Yeah, exactly! I personally enjoy playing Sudoku, so I spent quite a lot of time refining the puzzle generation and especially the difficulty design. I’m also unexpectedly happy with the UI/UX this time. I deliberately used Codex’s Build iOS plugin, and the result exceeded my expectations. AI-generated UI/UX is usually not that strong within codex, but this time the native SwiftUI version turned out surprisingly polished.

@BubuStd codexで作ったんだ、いいな。数独ゲームって意外とコード量あったりするけど、きれいにまとまってるのはセンスあると思うw

Shipped a clean sudoku game. Build with codex. I love it

Shipped a iOS local model personal assistant app. Build with codex.

playable link: bubucn.com/en/ai-model-ev…

Draft by kimi k3, fine tune with gpt 5.6 sol, its fun to play though

just under 30s, to finish the round. quick reflax?

居然还有个总使用量, 那么有没有什么说法, kimi vs kimi code消耗量区别?

Build with Kimi K3 with Swarm enabled. I'm now satisfied with its ability to generate 3D game demos. #kimi #kimik3

I am wondering how kimi-k3 high and low's performance.

kimi k3 one shot the blade run game, looks decent. #kimi #kimik3

yep, there are monsters.

Kimi K3 one shot backroom lv0 game #kimi #kimik3 #vibecoding #opensource

Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.com/blog/kimi-k3

I’m in. Cursor Pro plan just for Grok 4.5. It’s fast and efficient. @elonmusk My referral link btw: cursor.com/referral?code=…

Impressive, and which plan should I go for?

We've partnered with SpaceXAI to train Grok 4.5. It’s our most powerful model yet and the first we've built for more than software engineering.