Friday, August 14, 2026
Today's brief covers the latest llm-gemini 0.33 release enabling terminal-based Gemini 3.7 reasoning alongside Hugging Face's analysis of open-weight licensing and ecosystem shifts.
In this issue · 8
Writer has introduced its flagship Palmyra X6 model along with significant upgrades to its agentic harness. The company estimates the new model combined with harness changes will cut costs for customers by as much as 50% for basic tasks, while research found harness optimizations reduce costs by an average of 40%.
Google has launched Sheets canvas, a dynamic visual layer for Google Sheets built on Gemini. Teams can instantly transform raw row and column data into interactive dashboards, trackers, and seating charts using natural language prompts.
DeepSeek has updated its API pricing structure, introducing differential rates for peak and off-peak demand periods alongside revised calculations for prompt cache hits. AI developers should adjust scheduled execution windows to optimize API expenditures.
Microsoft is merging personal Microsoft Copilot and Microsoft 365 Copilot into a single unified application across desktop, web, and mobile. The update streamlines workspace clutter while retiring legacy features including Deep Research and Group Chat.
Vault Operator embeds an AI agent directly into Obsidian notes, providing block-level provenance, semantic search, and long-term memory across sessions. It also runs as a Model Context Protocol server that exposes your knowledge graph to external AI tools.
A practical guide details how to configure `mcp-confluent` to manage Apache Kafka clusters via natural language in Claude Code. It covers setting up SASL/SCRAM and mTLS authentication, read-only guards, and known REST limitations.
Simon Willison released version 0.33 of the llm-gemini plugin, introducing support for Google's Gemini 3.7 Flash and Gemini 3.5 Flash-Lite. The release enables CLI developers to inspect reasoning traces and run server-side code execution tools directly from the terminal.
Hugging Face published its Summer 2026 report analyzing nearly 3 million open model repositories. The data highlights Chinese labs dominating frontier sizes up to 2.78T parameters under permissive MIT/Apache licenses, while US contributions concentrate in hardware enablement.
Email digest
One email a day — the stories that matter for engineers, founders and tech leads. Human-edited, with links to primary sources.
By subscribing you agree to the privacy policy.