TLDR AI 2026-11-05

OpenAI is rolling out Images 2.6 globally in ChatGPT and Codex. The new model generates images 50% faster than the previous iteration, and it can also create GIFs. Users can adjust the speed of the GIF, edit individual frames, and share it. GPT-Image-2.6 Flare and Sunburst are now available through the API.

Ripple-65B Preview is a frontier open-weight model that achieves state-of-the-art performance on TerminalBench 2.1. It surpasses significantly larger open-weight models from the US, including Reflection Beam 501B, Meta Muse Spark Open 816B, Thinking Machines Inkling 975B, and Nvidia Nemotron 3 Ultra 550B. The model was trained for only $32 million, one-tenth the estimated cost of the previous open-weight SOTA on agentic coding. Ripple-65B Preview is available on Hugging Face, Together AI, GMI Cloud, and OpenRouter.

Pangram Image 1.0 is an AI image and AI video detection model that can produce a heatmap localizing exactly where AI appears in images. It is robust to edits, compression, and low resolutions. There is a Chrome Extension available that can detect AI, as well as API support for images and video.

OpenAI developed an autonomous pipeline for discovering, exploiting, and patching security vulnerabilities. In a single run, its cybersecurity model validated and generated patches for 948 Linux local privilege escalation vulnerabilities.

Axiom has released a Lean formalization of the entire Stacks Project. The Stacks project is an open-source reference that lays the foundation of modern algebraic geometry. The release is the largest integrated, reusable library of formalized mathematics ever built. It makes the full mathematical development of the Stacks Project available as a single, machine-checked library that researchers and AI agents can import from directly into new formal proofs.

Every new monitoring tool, service, or "quick integration" adds another dashboard to the pile—and before you know it, you're spending more time switching between tools than actually building. With Microsoft Azure, your entire stack lives in one workspace: infrastructure, data, AI, and more. See how Azure brings it all together →

SWE-2 Flash is a coding model that uses a new mid-training stage before reinforcement learning.

Hy4 is a large language model with visual understanding capabilities. Tencent plans to release model weights and system cards for the model in the coming days. The model supports a broad range of tasks that include fine-grained image processing, complex chart analysis, and mathematical visual reasoning. It can interact with both command-line interfaces and GUIs to perform a wide range of work and everyday tasks.

Cartesia's H-Nets learn directly from raw bytes, matching benchmark performance with substantially less training compute.

AlphaEvolve is a Gemini-powered coding agent for designing advanced algorithms. It pairs the creative problem-solving capabilities of Gemini models with automated evaluators that verify answers. The agent has been used to enhance the efficiency of Google's data centers, chip design, and AI training process. This repository contains a minimal implementation of AlphaEvolve along with code and data related to solutions and discoveries it has contributed to.

OpenAI will acquire Pierre Computer Company to enhance Codex's developer tools, including Code Storage and other open-source projects.

Conductor is a startup that lets developers run multiple AI coding agents in parallel workspaces.

Most vendors police AI agents at runtime. Virtru's Data Harbor Platform binds policy to the data. No entitlement, no key, no data. Launches live online, Nov 17.

Register now

Researchers find that frontier language models perform well with shorter contexts without sacrificing accuracy.

Together Compiler is a truly continual optimization loop that leverages high-throughput agents to optimize for diverse inference workloads.

Perplexity's Decisions API can now search the web before it answers.

Cat Wu has joined OpenAI to build reliable agents that keep users in control.

Kimi K3.1 Pro, the latest in the 3T-class models, excels in multimodal tasks, improving outputs in coding, 3D creation, and factual accuracy with a million-token context window.