
OpenAI updates the Model Spec with five values
OpenAI consolidates five values in the Model Spec and adds guidance on context and supervision; models still do not fully reflect this goal.
252 artigos por Marcelo Santiago

OpenAI consolidates five values in the Model Spec and adds guidance on context and supervision; models still do not fully reflect this goal.

Mistral has opened the weights for Large 4, with reported gains in coding and cyber defense. See benchmarks, limits and execution options.

Epic's VP of development described faster work and more idea exploration with AI in Unreal Engine. The remarks do not announce a public feature.

Google has taken AI Mode ads to ten new countries and plans to reach Brazil, Mexico and the United Kingdom by the end of 2026. See what has already been announced.

Akai teasers show vibe coding mods on the MPC Sample Key. The CEO talks about early access and open software, but there is no license or SDK yet.

The Wiz interview adds a warning about AI and PHP2Shell. It does not announce a new flaw: the focus is on team response.

ServiceNow released an MCP plugin for ChatGPT; security, voice and Codex in AI Workflow Factory are still on future roadmaps.

Anthropic added shareable legal artifacts to Claude, with more than 60 legal connectors and final review by attorneys.

Anthropic has rolled out smart reports, in-app surveys and shared budgets for teams using Claude Enterprise.

GitHub unveiled Projects, Fusion, session continuity and automations to coordinate agents from issue to merge in Copilot.

Chrome suggests flight, order and tracking data found in Gmail when filling out forms, with controls in the browser.

Google opened a preview of the Connected Apps Platform to integrate apps with Search and Gemini via MCP or AppFunctions.
In-depth technical curation: agentic architectures, real token benchmarks, MCP servers and open-source tools in your inbox and on the blog.