BREAKING 🚨: GOOGLE UPGRADED GEMINI 3 DEEP THINK! IT ACHIEVED SOTA SCORE OF 84.6% ON ARC-AGI-2 BENCHMARK.

GEMINI IS BACK! 👀

Upgraded Gemini 3 Deep Think performs at a gold medal level in Physics and Chemistry.

It is now available to Google AI Ultra Subscribers and the Gemini API via the Early Access Program opt-in.

BREAKING 🚨: OpenAI released GPT-5.3-Codex-Spark, a new, faster model powered by Cerebras infrastructure!

Available to Pro users only as a research preview.

No apps 👀

Google rolled out Design System support for Stitch, where users can create a set of design tokens to be consistently applied across generations.

Cline drops CLI 2.0 coding agent, powered by K2.5 and M2.5 for free

Cline CLI 2.0 brings its AI coding agent to the terminal with both interactive and autonomous modes, ACP editor support, pipeline-ready automation features, and a limited-time trial using Kimi K2.5 and MiniMax 2.5 via Moonshot AI.

🗞 #ai

Cline released Cline CLI 2.0, an open-source AI coding agent powered by Kimi K2.5 and MiniMax M2.5, available for free!

npm install -g cline on Windows, Mac, and Linux.

Open-source strikes back 👀

BREAKING 🚨: MANUS AI IS ROLLING OUT A NEW ALWAYS-ON AGENT FUNCTIONALITY!

SKILLS, SUBAGENTS, MEMORY, DEDICATED COMPUTER INSTANCE, IDENTITY AND MESSENGERS SUPPORT!

Meta is coming for OpenClaw 👀

ICYMI: Google opens Gemini 3 Deep Think API in Early Access

Google has updated Gemini 3 Deep Think, a reasoning tool in its Gemini app and API, aimed at researchers and engineers solving complex problems in science and engineering. It now covers more disciplines and supports new workflows via API integration.

🗞 #gemini

ICYMI: OpenAI debuts Codex-Spark powered by Cerebras infra

OpenAI released GPT-5.3-Codex-Spark, a model focused on real-time coding with near-instant feedback, optimized for rapid iteration on low-latency hardware. It's available to ChatGPT Pro users, with limited API access during early rollout.

🗞 #chatgpt

ICYMI: Z.AI launched GLM-5, new open-source model on chat and APIs

Z ai released GLM-5, an open-weight LLM for complex engineering and long-horizon tasks, featuring 744B parameters, DeepSeek Sparse Attention, and top-tier coding benchmarks, aimed at full-project execution and sustained context handling.

🗞 #ai

Manus AI launched 24/7 Agent via Telegram and got suspended

Manus AI launched Agents with persistent memory and custom skills, integrated via messengers like Telegram. Telegram support was suspended post-launch, limiting expansion. Pricing and broader platform access remain key challenges.

🗞 #manus

ICYMI: MiniMax debuts MiniMax-M2.5 model on web and APIs

MiniMax has launched its M2.5 language model for developers and enterprises, focusing on coding, automation, and office tasks. It supports multiple languages, delivers top benchmark results, and offers lower operational costs via API access.

🗞 #minimax