Search “Agents”

Anthropic shares when to fold subagents into the main agent loop

In comparisons across multiple enterprise deployments, Anthropic's Applied AI team found that a single agent with skills often outperforms subagent architectures in quality, with lower per-task cost and latency, and presents a migration method for folding subagents into the main loop. The article notes that subagents remain suitable for scenarios where context bloat degrades performance, independent tasks need to be processed in parallel, or tool selection conflicts with instructions, and recommends first treating subagents as tool calls, then converting them one by one into skills and MCP tools, using end-to-end evaluations to compare performance before and after. The article mentions that Claude Code's system prompt for more advanced models has been reduced by over 80%, accompanied by Claude API skills such as build-eval, prompt-audit, hillclimb, and cost-optimize.

Who Is On Call After an AI Agent Goes Live: Nine Engineering Steps for Production Ownership

The New Stack published an article discussing the ownership problem for production-environment AI agents, pointing out that teams often focus their energy on prompts, retrieval, tools, and model selection while overlooking who is responsible for the system. The article lays out nine actionable steps, including defining an owner and escalation channel for the agent, using a tool allowlist that denies unauthorized operations by default, setting budgets for execution steps and tool calls, logging model calls, distinguishing failures from escalations, requiring human approval for high-risk operations, providing a health check endpoint, blocking by default when security controls fail, and creating a degraded-operation runbook for middle-of-the-night incidents.

DeepSeek Search ranks 17th on the Artificial Analysis Search Index

DeepSeek Search scores 69 on the Artificial Analysis Search Index, ranking 17th among 27 search products, and is the lowest-priced paid search API on the list. It charges $0.75 per 1k searches, with a per-task cost of about $0.032, ranking 5th cheapest among 27 products. On AA-Omniscience its accuracy is 70%, ranking 10th, but on BrowseComp, which requires multi-hop reasoning, it is only 63%, ranking 25th; each search returns about 7k tokens, raising model cost to the 2nd highest on the list.

Fireworks Offers Free Training of Muse Glimmer 30B: Using SFT to Predict Kalshi Market Outcomes

Fireworks AI announced that Muse Glimmer 30B can be trained with SFT for free until November 30, and released a complete recipe for a prediction experiment using historical Kalshi market data. The fine-tuned Glimmer achieved 58.3% accuracy on unseen topics, 11.2 percentage points higher than Jev's 47.1%; in a historical backtest, a hypothetical $1,000 grew to $1,071 after accounting for simulated fees, while both base Glimmer and Jev lost money. Training used LoRA adapters and a K-way cross-entropy loss, ran through the training data only once, and separately used a calibration set to fit a temperature parameter for each question category.

English translation is being prepared.

English translation is being prepared.

Developers use AI reverse engineering to run iOS 16 on Android phones such as Redmi

Developer dipa-studio successfully booted and ran iOS 16.7.16 on a Redmi Note 10 Pro, able to enter the home screen, operate Settings and Control Center, and subsequently similar demonstrations appeared on devices such as the Redmi Note 12, Nothing Phone, and OnePlus 12. The porting process extensively used Claude and Codex to write driver adaptation code, and a reverse-engineering tool REA that automatically analyzes hardware interfaces also emerged. In the current version, cellular network, Bluetooth, camera, and speaker are all unavailable; GPU hardware acceleration has not been made to work, causing lag; booting takes 4 minutes; code signing verification and sandbox restrictions have also been relaxed; it is only suitable for early-adopter demonstrations.

English translation is being prepared.

Google Developers Blog publishes ADK voice agent tutorial: using streaming tools to eliminate silence

Google Developers Blog published a step-by-step tutorial on how to build a voice agent with ADK and gemini-3.8-live that does not stay silent for long periods. The core approach is to change tool functions from return to yield to make them streaming tools, or to use FunctionTool's response_scheduling to deliver results when idle; in the sales report benchmark flow, the agent's time to first speech dropped from 14.4 seconds to 1.2 seconds. The tutorial also covers the stop_streaming cancellation mechanism, preventing duplicate calls, and the trade-offs among the three scheduling modes WHEN_IDLE, INTERRUPT, and SILENT. It requires google-adk >= 2.10.0 and uses run_live bidirectional sessions.

Artificial Analysis launches HealthBench and HealthBench…

Artificial Analysis has launched leaderboards for OpenAI's two health dialogue benchmarks, HealthBench and HealthBench Professional; the former contains 5,000 conversations with general users and health workers, while the latter consists of 525 conversations all aimed at clinicians. On HealthBench Professional, GPT-6 Astra (max) ranks first with a length-adjusted score of 66.0%, GPT-6.1 Sol (max) is at 65.1%, and Claude Sonnet 5.5 (max with fallback) is at 63.0%. On HealthBench, Claude Sonnet 5.5 (max with fallback) leads at 62.1%, while DeepSeek V4.1 Flash (max) is at 60.9%; Claude Opus 5.5 and Claude Fable 5.1 average about 5,900 characters per answer and lose more than 11 points after length adjustment.

How to run n8n AI workflows on NVIDIA DGX Spark without calling cloud APIs

n8n Blog published a tutorial on how to deploy a complete model stack locally on a single NVIDIA DGX Spark, enabling n8n workflows to perform transcription, OCR, document Q&A, and image generation without calling cloud APIs. The entire stack exposes an OpenAI-compatible endpoint through LiteLLM; the main model is the NVFP4 version of Qwen3.6-35B-A3B, with a single-user generation speed of 120 tokens/s and an aggregate of 359 tokens/s across eight users. The author's testing found that image generation reduces chat speed to about one-third, and recommends that small teams queue image tasks or schedule them outside working hours.

TinyFish Makes Search Free: Breaking Down the Real Cost of AI Agents Reading Web Pages

TinyFish announced that, within limits, it will waive retrieval fees for Search and Fetch: Search is limited to 30 requests per minute and 500 per hour, while Fetch is limited to 150 URLs per minute and 1,000 per day. The article notes that free retrieval does not eliminate the inference costs of downstream models; the same article of about 4,300 characters returns an estimated 1,184 to 42,774 tokens across different providers, and three pages of unfiltered content can fill a 128K context window. In the September 28, 2026 Artificial Analysis snapshot, TinyFish Search scored 71 at $9.30 per 1,000 tasks, the lowest total cost among the displayed configurations but also the lowest score, with about 59 seconds per task.

English translation is being prepared.

English translation is being prepared.

Arena Alignment Index Update: Three New Models from Meta, Reflection, and StepFun Join the Leaderboard

The Arena Alignment Index has been updated, adding two new labs and three new models. Meta's Muse Wonder ranks 6th at 77.5%, the highest score among the three new models, with an unauthorized action rate of 2.09% and a deceptive completion rate of 14.41%, both the lowest of the three; Reflection's Beam-1.0-501B-A23B ranks 7th at 75.3%, with a false attribution rate of 4.50%, the best of the three; StepFun's Step 5 Preview ranks 8th at 75.1%, with an unauthorized action rate of 2.11%, close to that of Muse Wonder.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.

English translation is being prepared.