Making the final days of 2025 count Here’s everything we dropped this week: — Gemini 3 Flash, bringing fast frontier intelligence at a fraction of the cost to surfaces like @GeminiApp
, AI Mode in Google Search, @GoogleAIStudio
, Vertex AI, and more — CC, a @GoogleLabs
LLMS
-
Google Launches Gemini 3 Flash for Fast Frontier AI
By
–
-
Mixture of Experts Models: Current Perspectives and Efficiency
By
–
Poll: How Are You Thinking About Mixture of Experts (MoE) Models Today? MoE architectures scale model intelligence by routing each request to a subset of specialized “experts,” improving efficiency. Which best matches your current perspective?
-

Gemini 3 Pro Update and Performance Advancements Expected
By
–

BREAKING : Gemini 3 Pro is expected to get another update soon with advancements that enabled Flash to have a comparable performance. General Availability before NY?
-
OpenAI Adopts SKILLs Standard for Codex Task Configuration
By
–
OpenAI is adopting the SKILLs standard in Codex, introduced by Anthropic recently, to configure specific and reusable task instructions.
— 🚨 AI News | TestingCatalog (@testingcatalog) 19 décembre 2025
– ✅ MCPs
– ✅ SKILLs https://t.co/nsI6AI6lLv pic.twitter.com/Rf6lISeViEOpenAI is adopting the SKILLs standard in Codex, introduced by Anthropic recently, to configure specific and reusable task instructions. – MCPs
– SKILLs -
Codex Skills Support: Automating Linear Tickets and GitHub CI Fixes
By
–
Excited for Codex to support skills!
— Romain Huet (@romainhuet) 19 décembre 2025
This thread shares examples of skills we’re shipping, inspired by how we use Codex internally at OpenAI, like having Codex automatically update Linear tickets or fix GitHub CI failures. https://t.co/f9WWP9GMf2Excited for Codex to support skills! This thread shares examples of skills we’re shipping, inspired by how we use Codex internally at OpenAI, like having Codex automatically update Linear tickets or fix GitHub CI failures.
-
Querying Gemini Live integration with Notebooks
By
–
Wondering if you would be able to speak to your Notebooks via Gemini Live as well. Nice addition!
-

ODB-dLLM: Faster Parallelizable Diffusion-Based Language Models
By
–
Diffusion-based LLMs are fast and parallelizable, but bidirectional attention makes inference expensive due to repeated prefill and decoding. Enter ODB-dLLM, a dual-boundary framework with adaptive prefill length prediction and dLLM-specific jump-share speculative decoding.
-

9 AI Skills That Will Matter in 2026 Beyond ChatGPT
By
–
9 AI skills that will actually matter in 2026
(not just “use ChatGPT”) Prompt engineering RAG pipelines Agentic workflows Fine-tuning & distillation LLM evaluation & red-teaming Cost & latency optimization Tool calling + function calling Multimodal -

GPT 6 Is Coming to Crush Gemini3 Competition
By
–
GPT 6 is coming… And it's going to change everything Here's everything you need to know → https://
youtu.be/Kb61cQMlbGM The goal is clear: crush the competition, #Gemini3 in the lead! -

SonicMoE: Tri Dao Advances MoE Kernel Efficiency on H100s
By
–
The author of FlashAttention, Tri Dao, just dropped a new paper called SonicMoE With 1.86x higher MoE kernel throughput and 45% lower activation memory per layer on H100s, by introduceing tile-aware routing that cuts padding waste for sparse MoEs Trending on alphaXiv