252 artigos por Marcelo Santiago
Claude API 2026-10-21 rejects temperature, top_p and top_k with a 400 error. Which SDKs send the new version by default and how to migrate safely.
OpenAI is bringing its models and Codex workflows to Google Cloud's Model Garden. The preview starts with selected enterprise customers.
xAI introduced The X Subscription, a plan that bundles Grok, Grok Bot, Cursor and X Premium across five tiers. See pricing, shared usage, extra costs and how to evaluate migration for your vibe coding workflow.
CoreWeave test published by NVIDIA reports sandbox startup over 3x faster and agent loops 1.8x faster with Vera versus x86.
Google announced Gemini search in preview on isolated GDC. Answers cite source files and processing remains inside the local perimeter.
Microsoft added Retrieve to Work IQ MCP so agents can search authorized context in emails, files, meetings and Teams messages, with citations.
Anthropic has moved inference hooks out of beta for Claude Enterprise, allowing requests to be evaluated on a security server before they reach the model.
OpenAI acquired the Orchestra Bio platform and will merge its team and technology with Rosalind to coordinate drug research and development.
OpenAI released 12 hour prompt cache retention in the Responses API. See when it helps and the cost of writing to the cache.
Accenture Federal Services and Google Public Sector announced Gemini Launchpads to build and scale AI agents for federal government missions.
Microsoft proposes Guardian Agents to monitor AI agents and treat autonomous systems as new users within corporate security.
Reflection released Beam, an open weights model with an Apache 2.0 license built for code, reasoning and agents, with support for vLLM and SGLang from day one and its own terminal agent, Mirror CLI.
In-depth technical curation: agentic architectures, real token benchmarks, MCP servers and open-source tools in your inbox and on the blog.
