AI News | jls's blog

Claude Code switches to Auto Mode by default, Astra classified as critical in cybersecurity by OpenAI, NVIDIA unveils Cosmos 3

On August 7, 2026, Claude Code adopts Auto Mode as the default permission mode starting August 14, OpenAI classifies its future Astra model at the critical level for cybersecurity, and NVIDIA launches Cosmos 3, a family of open models for robotics and autonomous vehicles. The week also adds Claude Code inter-session messaging, a loosened biology policy for Claude Fable 5, general availability of Seedance 2.5, and new tools on the GitHub Copilot and agentic coding tools side.

August 7, 2026 Read more

claudeclaude-codeanthropic +11

GPT-5.6 Sol unifies ChatGPT, Kimi K3 reaches general availability in Copilot, WeatherNext predicts cyclones five days ahead

On August 6, 2026, OpenAI unifies ChatGPT around GPT-5.6 Sol with 68% fewer factual errors and makes GPT-5.6 Luna unlimited and free, Kimi K3 reaches general availability in GitHub Copilot, and Google DeepMind releases WeatherNext, an open model that forecasts cyclones five days ahead. Wan3.0 (Alibaba) enters public beta, Meta publishes Artificial Analysis benchmarks for Muse Spark 1.2 and claims five gold-level results at international science Olympiads, along with fifteen other announcements on agentic coding tools and the open-model landscape.

August 6, 2026 Read more

openaichatgptgpt +11

Meta launches Muse Code, OpenAI and Anthropic detail cybersecurity incidents, Zed secures its agent

August 5, 2026: Meta launches Muse Code powered by Muse Spark 1.2, OpenAI and Anthropic detail cybersecurity incidents that occurred during evaluations conducted by the UK AI Security Institute, and Zed defaults to sandboxing its agent tools in version 1.14.

August 5, 2026 Read more

metaopenaianthropic +6

Warp launches its Agent CLI, GitHub retires Spark, FLUX 3 Video available: the August 4 announcements

August 4, 2026: Warp launches an autonomous CLI agent, GitHub schedules the retirement of Spark, Black Forest Labs releases FLUX 3 Video, NVIDIA unveils Alpamayo 2 Super, and Mistral launches Shieldstral, its moderation model.

August 4, 2026 Read more

coding-toolsgithubcopilot +5

Qwen3.8-Max at Alibaba, GenOffice open source, API v0: the announcements of August 3

August 3, 2026: Alibaba unveils Qwen3.8-Max, Genspark opens the source code of GenOffice, Vercel launches a programmable API for v0, and Cursor, GitHub, Sakana AI, MiniMax H3, and Runway publish several notable updates.

August 3, 2026 Read more

qwenalibabagithub +7

Quiet AI weekend: Amp removes its leaderboard, Cohere Labs number 1 on Twitch

The weekend of August 1 and 2, 2026 was quiet on the AI announcements front: Amp removes the ranking page from its code agents tool, and Cohere says its Twitch streams are number 1 in the tech category.

August 2, 2026 Read more

actualitesiacoding-tools +4

MiniMax launches H3 with announced open weights, Together AI claims 10,000x token growth, GitHub locks down risky npm tokens

August 1, 2026: MiniMax launches H3, a multimodal model with announced open weights and native stereo audio, Together AI claims more than 10,000x growth in served tokens, and GitHub restricts npm tokens to 2FA bypass after a wave of account takeovers.

August 1, 2026 Read more

actualitesiaia-generative +22

Anthropic reveals three cybersecurity incidents, DeepSeek relaunches its API, Copilot gets a makeover in VS Code

July 31, 2026: Anthropic details three real incidents discovered in its cybersecurity evaluations, DeepSeek relaunches its API with V4-Flash-0731, and GitHub Copilot in VS Code wraps up July with a redesign of the Agents window.

July 31, 2026 Read more

actualitesiaanthropic +19

Claude discovers novel cryptographic flaws, Gemini Robotics ER 2 gets multiple robots collaborating, GPT-5.6 lowers its prices

On July 30, 2026, Claude Mythos reveals novel cryptographic flaws at Anthropic, Google DeepMind launches Gemini Robotics ER 2 for robot collaboration, and OpenAI cuts GPT-5.6 prices.

July 30, 2026 Read more

actualitesiaanthropic +17

OpenAI and Anthropic call for slowing down AI research, two rivals open-source code agent security, GPT-5.6 Sol optimizes its own infrastructure

On July 29, 2026, OpenAI and Anthropic jointly support a call to slow down research in artificial intelligence, two competitors release open source tools to secure code agents, and GPT-5.6 Sol optimizes its own service infrastructure.

July 29, 2026 Read more

actualitesiaopenai +17

MCP shifts to stateless architecture, Hugging Face details the July cyberattack, Perplexity Computer arrives on Windows

On July 28, 2026, MCP moves to a stateless architecture in its 5th major version, Hugging Face publishes the full timeline of the July cyberattack, and Perplexity Computer arrives on Windows with orchestration of 15+ models.

July 28, 2026 Read more

actualitesiamcp +16

Moonshot launches Kimi K3 (open 2.8T), NVIDIA announces Open Secure AI Alliance, Perplexity beefs up its enterprise offering

July 27, 2026: Moonshot AI releases Kimi K3, an open 2.8-trillion-parameter model; NVIDIA brings together an industry alliance on AI system security with more than 30 partners; and Perplexity publishes a dense enterprise changelog.

July 27, 2026 Read more

actualitesiakimi +12