List of Flash News about Tool Use
| Time | Details |
|---|---|
|
2026-02-03 06:54 UTC |
Google Gemini Interactions API Unifies LLMs and Agents: 3 Core Capabilities for Stateful Workflows and Tool Use
According to @rseroter, Gemini’s new Interactions API shifts beyond simple request-response to a unified, stateful interface spanning LLM calls and agent capabilities like Gemini Deep Research, with built-in tool invocation, source: @rseroter. The update emphasizes that a single API can manage stateful interactions, interact with agents, and invoke tools, enabling cohesive agentic workflows, source: @rseroter. A technical deep dive by @meteatamel describes the Interactions API as one interface for models and agents, highlighting the move toward integrated agents and tools under one endpoint, source: @meteatamel. |
|
2025-12-15 18:00 UTC |
Anthropic Claude Opus 4.5 cuts per-token cost to about one-third and boosts long-context reasoning and tool use, according to DeepLearning.AI
According to DeepLearning.AI, Anthropic’s new flagship Claude Opus 4.5 improves coding, tool use, and long-context reasoning while costing about one-third per token versus its predecessor, directly lowering unit inference costs relative to earlier Claude models (source: DeepLearning.AI on X, Dec 16, 2025; more details: hubs.la/Q03Yf3f60). It adds adjustable effort and extended thinking plus automatic long-chat summarization, features designed to manage reasoning depth and summarize lengthy interactions at lower token consumption than before (source: DeepLearning.AI on X, Dec 16, 2025). Independent benchmarks cited by DeepLearning.AI place Opus 4.5 near the top, and it often achieves comparable results with far fewer tokens, improving cost efficiency for long-context tasks compared with its predecessor (source: DeepLearning.AI on X, Dec 16, 2025). |
|
2025-12-11 09:47 UTC |
Andrew Ng Releases Open-Source aisuite: Build Highly Autonomous but 'Very Unreliable' AI Agents for Frontier LLMs with Tool Use
According to @AndrewYNg, the open-source aisuite package provides a recipe to build a highly autonomous, moderately capable, and very unreliable AI agent, enabling a frontier LLM to use tools such as disk access and web search with just a few lines of code; source: @AndrewYNg on X, Dec 11, 2025. For crypto market participants and quant developers, this signals experimental-stage agent tooling rather than production-grade automation, warranting caution in any live trading or custody context; source: @AndrewYNg on X, Dec 11, 2025. Ng also states he is collaborating with Rohit Prasad on aisuite, indicating active open-source development of agent tooling; source: @AndrewYNg on X, Dec 11, 2025. |