
OpenCode fixes GPT-6.2 prompt caching
OpenCode reports that GPT-6.2 now caches the system prompt from the second step on Responses routes.

OpenCode reports that GPT-6.2 now caches the system prompt from the second step on Responses routes.

OpenAI's Dominik Kundel and Charlie Guo showed how Codex can edit video and connect agent workflows at CODE Summit 2026.

Versions 1.26.1 and 1.26.2 of the agent-framework PyPI package were compromised to steal Azure, GitHub Actions, and other CI/CD credentials.

Grok Bot 0.70.2 or later lets you set each tool to Always allow, Auto-review, Needs approval, or Blocked.

Orca keeps your draft, reading position, and editor state saved in paused chats on macOS and Linux.

NVIDIA Vera Rubin NVL72 shows high agentic inference performance in MLPerf Endpoints v1.0 results.

Together AI says Together Compiler beats its hand-written megakernels by 28% on B200 GPUs on Llama 3.2 1B.

Developers can now run Claude Managed Agents, including the control plane, in their own cloud or on-premise environment.

AWS explains how to route coding agent requests to the right model on Amazon Bedrock using NVIDIA NeMo Switchyard and LiteLLM.

ChatGPT Wallet lets agents pay without access to users' actual card details.

For prompts up to 200k tokens, Muse Ember 1.0 is priced at $0.10 per million input tokens.

SolveIt costs nothing to join, and users pay only for what they use.

Hy4 flash supports multimodal inputs and a context length of up to 1 million tokens, Tencent says.

GPT-6.2 Astra’s API price drops 20% from GPT-6 Astra, with cached input at $0.40 per million tokens.

Langfuse introduced its AI Gateway in public beta, offering a central entry point to manage and trace all AI usage.

Runta says 64.5% of the list-price cost of the scored runs went to trials that did not pass.

OpenAI launched a partnership with the Chilean government to support responsible AI use in public services and boost local AI startups.

Alibaba's Qwen4-Max is a reasoning model for coding and long-running agent tasks.

Google launched the Gemini Decisions API for fast, probabilistic choices across text, images, audio, video, and PDFs using the Gemini Decision 1 model.

Google says users will see more visual content like tables and custom tools.
Sign in to see your saved feeds. Saving a custom feed is part of Full Access.
The builder's brief, in your inbox every weekday.
Free. Unsubscribe in one click.